v1.80.0.0 fix: setup survives a failed Chromium install, hooks share one state root, gstack never clobbers a skill it did not create (#2802)

* fix(freeze): hook reads the same state root /freeze writes — fails closed under GSTACK_HOME (#1459, #1509)

check-freeze.sh resolved its state dir as ${CLAUDE_PLUGIN_DATA:-$HOME/.gstack}
while every writer (/freeze, /guard, /unfreeze, /investigate) resolves through
bin/gstack-paths, GSTACK_HOME first. With GSTACK_HOME set, /freeze wrote
freeze-dir.txt under GSTACK_HOME, the hook read $HOME/.gstack, found no file,
and allowed everything — a deny-tier boundary failing open.

One resolver now: gstack_hook_state_root() in careful/bin/hook-extract.sh
(already sourced by both check-freeze.sh and check-careful.sh) implements the
exact gstack-paths chain, including the CLAUDE_PLUGIN_ROOT guard that keeps a
CLAUDE_PLUGIN_DATA leaked from another plugin from redirecting our state.
check-freeze.sh and gstack_hook_log_fire both call it; nothing spawns
gstack-paths from a hook.

Tests: the GSTACK_HOME deny regression, GSTACK_HOME-over-CLAUDE_PLUGIN_DATA
precedence, plugin-root guard both ways, and a byte-parity check against
bin/gstack-paths across six env combinations. Existing freeze tests now pass
CLAUDE_PLUGIN_ROOT like a real plugin install would.

Idea from PR #1509 (@NikhileshNanduri); implemented natively against the shared
resolver rather than a second fallback chain.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>

* fix(relink): never delete or link over a skill gstack does not own (#2119)

gstack-relink runs on every ./setup. Its cleanup did `rm -rf` on any same-name
entry whose SKILL.md was a symlink, with no readlink check, and its link step
did `mkdir -p` then `ln -snf` onto any existing SKILL.md — on Linux that
replaces a user's real file with a symlink into gstack (macOS refused by
accident). setup's Windows mode-flip cleanup deleted any real dir whose name
matched a gstack skill. A personal `qa` skill, or a fork installed under
another path, was destroyed by the installer of a tool it never asked for.

Ownership is now proven, never assumed. An entry is ours when it is a symlink
resolving into INSTALL_DIR or RENDER_DIR, a real dir whose SKILL.md is such a
symlink, or a real dir carrying the .gstack-owned marker setup now writes for
Windows copy installs (legacy copies count when byte-identical to the source
or carrying gen-skill-docs' AUTO-GENERATED header). Anything else — including
an entry whose readlink fails — is foreign: left untouched, reported on
stderr, and listed in relink's summary line. The same rule replaces setup's
Windows name-match deletion; setup:1040 and gstack-uninstall:204 already
gated on readlink, so this closes the last unguarded deleter of the class.

Tests: foreign real dir in flat mode, foreign flat entry on a prefix flip,
foreign directory symlink, RENDER_DIR-targeted entry (ours), marker-carrying
copy (ours), marker-less copy (foreign); the Windows cleanup test now proves
provenance three ways and keeps the user's own same-name skill.

Idea and two regression cases from PR #2119 (@smblight); implemented on the
destination entry, not only the symlink target.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>

* fix(setup): Chromium bootstrap is best-effort and bounded — skills always register (#1900, #1901, #1902, #913, #2233)

setup runs under `set -e`, and the Chromium bootstrap in section 2 sat ahead
of skill registration in section 4 with a bare `bunx playwright install
chromium`, an unbounded download, and an explicit `exit 1` after the
post-install launch probe. On an offline, proxied, or AppArmor-restricted box
the user ended with ZERO skills registered and a re-run that died at the same
line; a wedged download hung setup indefinitely.

Every browser failure now records a reason code in _PW_FAIL_REASON and setup
continues: skipped (GSTACK_SKIP_PLAYWRIGHT=1, #913), chromium-install,
chromium-install-timeout (the download is bounded by the existing
_wait_with_deadline helper, default 600s, env GSTACK_PLAYWRIGHT_INSTALL_TIMEOUT,
process tree killed via _kill_tree), chromium-install-locked (another setup
holds the lock: this one registers skills and re-probes next time instead of
exiting), windows-no-node, windows-node-modules, post-install-launch (with the
GSTACK_CHROMIUM_NO_SANDBOX=1 hint for Ubuntu 24.04's userns policy, #2157).
The daemon font refresh is skipped when Chromium is unavailable. The final
summary names the skills that need the browser (/qa, /qa-only,
/design-review, /browse, make-pdf, /pair-agent) and the fix for the recorded
reason, and logs the reason code (never a path) through gstack-telemetry-log
when telemetry is on.

Tests: static invariants over the anchor-sliced block (no exit, every reason
code, deadline helper, trap chaining, guarded refresh, summary contents) plus
an integration harness that executes the real block with a stubbed probe and
installer: install failure, hang killed at the deadline with the tree kill
recorded, non-numeric knob fallback, live lock (continues, installer not run,
lock preserved), stale lock reclaimed, post-install probe failure, and the
skip flag.

Credit @DavidMiserak (PR #1900) for the best-effort shape; re-implemented on
the current block.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>

* docs(designs): preserve the time-attack fork-port residual evaluation

The read-only evaluation of what remains portable from time-attack/gstack
(583 raw candidates, 415 canonical, 287 with a residual, 48 adversarially
refuted, 14 standing) lived only on a throwaway VM. This records the report,
the lite residual index, the absorbed/superseded ledger, the refuter
verdicts, and SHAS.md with the fork tip, upstream HEAD, merge-base, and a
sha256 per file, so every scheduled fix in this wave series traces to its
evidence.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>

* docs: file the fork-port residual deferrals and document the Chromium bootstrap knobs

TODOS.md gains the seven items the CEO and eng reviews of the fork-port
residual plan deliberately deferred (shared ownership helper, config-key
reader tripwire, "pre-existing" vocabulary, opt-in reply_language, .auth.json
writer removal, the fork-derived-change rule for CONTRIBUTING, hook slug
parity audit), each with rationale, and updates the two residual bullets for
PR #2232 and PR #2233 with their dispositions. README's Troubleshooting
section explains the best-effort Chromium bootstrap and its three knobs.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>

* fix(relink): canonicalize link targets before the ownership check

Pre-landing review finding: the ownership gate compared readlink output
textually against INSTALL_DIR and RENDER_DIR, so two shapes of gstack's OWN
entries read as foreign and were left behind on a mode flip — a legacy
relative link (`gstack/qa/SKILL.md`, resolved against $PWD instead of the
link's directory) and an entry linked against the real path of a symlinked
install dir (~/.claude/skills/gstack -> checkout). Both now resolve: relative
targets anchor at the link's directory, the directory part is canonicalized
with pwd -P (the basename stays verbatim so a dangling managed target is not
misread), and both spellings of each root are accepted. Two regression tests.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>

* fix(telemetry): one-shot setup events never sweep other sessions' pending markers

gstack-telemetry-log finalizes every .pending-<session> marker that is not
the caller's own as outcome:unknown and deletes it. setup's onboarding
events (_setup_welcome, _setup_playwright) have no session of their own, so
a Chromium bootstrap failure during a live skill session recorded a false
unknown for that session and removed its marker.

New --no-sweep flag skips the stale-marker pass; both setup call sites use
it (the synthetic --session-id did not prevent the sweep). Surfaced by the
Codex adversarial pass.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>

* fix(hooks): partial upgrades fail closed for freeze and fall back for careful

A hook script and its sourced helper can be copied at different times. With
an older careful/bin/hook-extract.sh that lacks gstack_hook_state_root:

- check-freeze.sh now emits a deny ("fail closed, re-run ./setup or
  /unfreeze") instead of dying under set -e with no decision JSON.
- check-careful.sh falls back to ${GSTACK_HOME:-$HOME/.gstack} so project
  rules under the plain chain still load and a decision is always emitted
  (a warn hook must never break on a stale helper).

gstack_hook_state_root prints its root without a trailing newline and both
callers capture it with a printf-x sentinel, so a GSTACK_HOME ending in a
newline round-trips byte-for-byte with the writer's %q form.
gstack_hook_log_fire stays on ${GSTACK_HOME:-$HOME/.gstack}/analytics, the
same two-step chain every other analytics writer and reader uses, so the
usage log remains one file under a plugin install.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>

* fix(setup): never link over, copy over, or reap a skill gstack does not own (#2119)

The relink gate alone left three destructive sites open:

- link_claude_skill_dirs runs BEFORE relink on every ./setup and used
  `ln -snf` (Linux replaces a user's real SKILL.md with a symlink into
  gstack) or, on Windows, rm -rf + cp followed by a marker that made the
  user's directory "ours" on the next flip. It and _install_alias_skill_md
  now consult _claude_entry_is_ours first and skip loudly.
- cleanup_prefixed_claude_symlinks kept a bare name-match deletion and a
  `*gstack*` substring match. Symlink arms use anchored `gstack/` segment
  patterns; the Windows real-file arm proves provenance (marker,
  byte-identity with our source, or the full two-line gen-skill-docs banner
  within the first 40 lines, never a one-line substring another generator
  could emit). cleanup_old_claude_symlinks uses the same banner rule.
- gstack-relink's fast path judged absolute targets before canonicalizing,
  so `/x/gstack/../foreign/SKILL.md` counted as ours; dot-segment targets
  now canonicalize first. Its banner rule matches setup's.

The `.gstack-owned` marker records the owning payload's realpath. Entries
skipped by setup or relink are listed in the final setup summary.

Chromium bootstrap refinements from the pre-landing review: an INT/TERM
trap kills the installer's process tree; the Windows npm chain no longer
masks an install failure; GSTACK_SKIP_PLAYWRIGHT=1 is reported as a choice
rather than a failure and sends no telemetry; the timeout knob is
normalized (0, 000, non-numeric, or more than nine digits fall back to the
600s default instead of killing on the first poll or never killing).

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>

* docs: README Chromium note outside the CLAUDE.md fence; report banner stripped; deferrals name the four gate sites

- README: the Chromium troubleshooting paragraph sat inside the CLAUDE.md
  snippet code fence, so copy-paste put it into users' CLAUDE.md. Moved to
  the troubleshooting list.
- docs/designs/fork-port-residual-2026-09/REPORT.md: the scratch-run
  preamble banner is gone; SHAS.md re-hashed.
- TODOS: the ownership-gate deferral names the four sites and the
  marker-path idea for the fork-with-banner residual.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>

* docs(todos): the bootstrap block coverage gap is pinned except the quarantine helper

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>

* fix(setup,relink): ownership proof has two strengths; weak proof never deletes a directory or discards a differing file

The first #2119 gate treated a byte-identical or banner-bearing real-file
SKILL.md as full ownership, so a prefix flip could rm -rf a user's directory
(their own qa skill started from a gstack SKILL.md, plus my-templates/) and
the link pass could replace their customized file with a symlink. Two
strengths now:

- STRONG: the .gstack-owned marker (we created the directory), or a
  directory holding nothing but symlinks and the marker (deleting it loses
  no data). Only strong proof removes a directory whole.
- WEAK: byte-identity with our source or the two-line gen-skill-docs banner
  on a real file. Weak proof covers that SKILL.md and our runtime-asset
  links only; a differing file is moved to
  ${GSTACK_HOME:-~/.gstack}/backups/skills/<ts>/<skill>/ before we link
  over it, and setup/relink print one summary line naming what moved.

The marker is written on every platform now (path-independent proof for
Windows copies and for checkouts whose path carries no gstack segment), but
only for a directory gstack creates: a directory we merely link into
(unclaimed, or a legacy install) never becomes deletable whole. A directory
with no SKILL.md at all is unclaimed: the link pass may add our file, the
cleanup pass has nothing to remove.

Also from the review passes: the banner check reads 8192 bytes, not 40
lines (investigate, office-hours, plan-ceo-review and design-consultation
carry the banner past line 40 and were left "foreign" on pre-marker
Windows installs); a link into a checkout named without a gstack segment
(git worktree add ../gstack-<branch>) is ours when that tree carries
setup + VERSION + bin/; relink's fast path is gone so both files
canonicalize before judging; relink's root alias (_gstack-command) is
gated and stamped like every other entry; relink reports the bare entry
name with setup's wording and setup dedupes when forwarding
(_run_relink_quiet); the summary names the browser skills as examples.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>

* fix(setup): Chromium-install lock reclaim is atomic and pid-validated; abandoned locks expire; the tree kill walks /proc without pgrep

- A pid file holding "", "-1" or "0" counted as a live holder (kill -0 -1
  signals every process and succeeds), locking Chromium out for good. A pid
  must be a positive integer; anything else is stale.
- Two setups judging the same lock stale raced on rm -rf + mkdir and the
  loser deleted the winner's fresh lock. The stale dir is renamed first
  (atomic), so exactly one reclaims.
- A lock dir with no pid file (killed between mkdir and echo) was never
  reclaimed; it now expires once older than the install bound.
- _kill_tree needed pgrep; debian-slim and git-bash ship none, so the bound
  killed only the wrapper subshell and the installer kept running. Without
  pgrep the children are found by walking /proc/*/stat.
- The timeout knob is normalized in one place with one comment; the trap's
  exit 130 is the only exit the block may contain.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>

* fix(freeze): an unexpected non-zero death denies via an EXIT backstop instead of exiting with no decision

set -e plus a failing pipeline (a tool on PATH exiting non-zero, a deleted
cwd) ended the deny-tier hook with no JSON, which Claude Code treats as
non-blocking: the edit outside the boundary proceeded. The EXIT trap now
prints a deny for any non-zero exit that happens before a decision was
written; every deliberate output sets _FREEZE_DECIDED first so a late
failure never prints a second object.

Tests also pin careful's state-root precedence (GSTACK_HOME over
CLAUDE_PLUGIN_DATA, plugin data when CLAUDE_PLUGIN_ROOT names gstack) and
the specific "out of date" deny for a helper without gstack_hook_state_root.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>

* refactor(telemetry): guard the stale-marker sweep with an if, not a break inside the loop

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>

* docs(todos): the ownership gate lives in six sites, and the cleanup arms inline their own chain

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>

* test: the two remaining linker harnesses extract the ownership helpers; the marker is the one allowed dotfile

setup-claude-skill-assets and user-render-out-dir-install slice
link_claude_skill_dirs out of setup without the helpers it now calls, so
the extracted function died with "command not found" (or, inside an if,
degraded into "foreign, skipped"). Both harnesses now carry the full helper
set and the globals. The hidden-files census allows .gstack-owned, which
the linker writes for directories it creates rather than copying from the
skill source.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>

* fix(setup,relink): weak proof never costs the user a file — assets, flips, failed backups, foreign dir links, alias markers

Third review cycle on the ownership model, every item reproduced against a
fixture before the fix:

- Runtime assets (sections/, templates/, checklist.md, ...) were refreshed
  with rm -rf regardless of who owned the directory, so an unclaimed or
  weakly-owned directory lost the user's same-named real files. Real assets
  are now replaced only in a directory gstack created or strongly owns
  (marker, or SKILL.md symlink into gstack), plus the legacy Windows
  real-copy shape; elsewhere they are kept and reported. Symlinks are never
  content and are always refreshed.
- The prefix-flip cleanup deleted a customized banner-bearing SKILL.md that
  the link pass would have backed up. Both cleanups now compare the file
  against the source (raw, or with its name: line rewritten to the entry
  name, which is how alias and prefixed copies legitimately differ) and
  move a differing file to the backup root.
- A failed backup (unwritable root) returned success and the caller linked
  over the file anyway. It now fails, and the entry is left untouched and
  reported.
- A foreign DIRECTORY symlink whose target had no SKILL.md fell through to
  the "unclaimed directory" rule and was replaced by a real directory. A
  symlink that does not resolve into gstack is foreign, full stop.
- The alias installers stamped .gstack-owned into pre-existing directories;
  they now follow the same created-or-already-marked rule.
- A directory counts as "only links" only when every link resolves into
  gstack: a user's own symlink makes it mixed, so their link survives.
- The gstack-tree heuristic requires bin/gstack-relink, not just a VERSION
  file, a setup script and a bin/ directory.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>

* fix(setup): lock reclaim hands a fresh lock back; a live holder past the bound is stale; /proc walk strips through the last paren

- Reclaim renamed the lock by path after judging it stale, so a second setup
  that had already reclaimed and re-created it lost its fresh lock and two
  installers ran. After the rename the moved directory's pid is re-read: a
  new live holder, or a fresh lock whose pid is not written yet, is moved
  straight back.
- A pid file whose process is alive but whose lock is older than the install
  bound is stale too (the holder is past its own deadline, or the pid was
  recycled to an unrelated long-lived process); it was locked forever.
- The /proc fallback stripped the comm field to the FIRST ") ", so a comm
  containing ") " hid a child from the kill. proc(5) says the last paren.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>

* fix(freeze): mark the decision written after the helper prints, not before

If gstack_hook_decision ever failed between the flag and its output the
backstop would have stayed silent; setting the flag after the print keeps
the deny backstop armed until a decision is actually on stdout.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>

* chore: bump version and changelog (v1.80.0.0)

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>

* docs: update project documentation for v1.80.0.0

README troubleshooting + manual uninstall cover the skill ownership gate
(.gstack-owned marker, ~/.gstack/backups/skills/<ts>/, foreign same-name
skills left untouched). CLAUDE.md and CONTRIBUTING carry the ownership and
best-effort Chromium bootstrap invariants for people editing setup and
gstack-relink. PROJECT_STRUCTURE gains careful/, freeze/, guard/, unfreeze/,
gstack-upgrade/, gstack-relink, and the setup/relink/hook test files.
TESTING_INTERNALS documents the anchor-sliced setup harness convention.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>

* fix(setup): the final summary reports customized SKILL.md files moved to the backup root

The linker moved a weakly-proven, customized SKILL.md aside before linking
over it but never said so; only relink printed a "Moved N" line, and by the
time relink runs the file is already a symlink. The summary now names each
moved file and where it went, next to the foreign-entry report.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>

* test: run assembled setup harness scripts from a temp file, not `bash -c` argv (Windows MSYS2 8 KB truncation)

windows-free-tests (run 33907177851) failed in
test/setup-alias-name-uniqueness.test.ts with
  bash: -c: line 178: unexpected EOF while looking for matching `'
The harness slices functions out of `setup` and passed the joined script as
one `bash -c` argv element. The ownership gate grew that script from 6.7 KB
to 15.7 KB, and on Windows bash is an MSYS2 program: when its parent is a
non-MSYS process (bun), msys-2.0.dll's build_argv() runs any argument
containing `?*["'(){}` through globify()/glob(), which copies the pattern
into a fixed `Char patbuf[8192]` and silently stops after 8192 - MB_CUR_MAX
(8186 chars under C.UTF-8); GLOB_NOCHECK then returns the truncated text as
the argument. Character 8186 lands inside the single-quoted sed token on
line 178. Rebuilding the exact script with CI path shapes and cutting it at
8186-8190 characters reproduces the identical message locally; cmd.exe's
8191-UTF-16 cap and CreateProcess's 32767 do not fit the evidence.

Fix: test/helpers/bash-script.ts writes the script to a temp file and runs
`bash <path>` — a short glob-free argument that never enters globify. Every
setup harness that assembled a script for `bash -c` (11 files, 22 sites)
uses it; timeouts and env are preserved verbatim, spawn/timeout errors are
appended to stderr, temp cleanup is best-effort. `spawnSync('bash',
[<Windows absolute path>])` already passes on windows-latest in setup-help,
uninstall-windows-copies and the migration tests. The Windows-curated list
is byte-identical before and after.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>

* docs(test-free-shards): the rerun-refresh harness spawns bash <tempfile> via test/helpers/bash-script.ts, not bash -c

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>

---------

Co-authored-by: Claude Fable 5.1 <noreply@anthropic.com>
This commit is contained in:
Garry Tan
2026-09-05 14:45:28 -07:00
committed by GitHub
co-authored by Claude Fable 5.1
parent 0d1bd5616c
commit c241216637
38 changed files with 5938 additions and 137 deletions
+454 -54
View File
@@ -116,6 +116,177 @@ _link_or_copy() {
fi
}
# ─── Ownership gate for skill entries (#2119) ─────────────────────────────────
# setup and gstack-relink must never delete or link over a skill they do not
# own. This is the single rule both use (relink carries the same logic — keep
# them in sync until the shared helper filed in TODOS.md lands). Proof has two
# strengths: STRONG (a symlink resolving into gstack, or the .gstack-owned
# marker) means we created the entry and may delete or refresh it whole; WEAK
# (byte-identity with our source, or the two-line generated banner on a real
# file) covers only that SKILL.md — never the directory — and a differing
# weakly-proven file is moved to ${GSTACK_HOME:-~/.gstack}/backups/skills/<ts>/
# before we install over it. An entry is OURS
# when: it is a symlink resolving into the gstack payload / render dir (or any
# path with a `gstack` segment, the convention cleanup and gstack-uninstall
# already use, so entries from a sibling worktree still count), a real dir
# whose SKILL.md is such a symlink, or a real-file copy proven by the
# .gstack-owned marker, byte-identity with the source, or gen-skill-docs'
# generated header. Anything else is FOREIGN: skipped, reported, listed in
# the final summary.
_FOREIGN_SKIPPED_ENTRIES=()
_gstack_link_target_abs() {
# readlink of a relative link is relative to the link's directory; anchor it
# there and canonicalize the directory part (`..`, symlinked components).
local link="$1" dest d b d_real
dest="$(readlink "$link" 2>/dev/null || true)"
[ -n "$dest" ] || return 1
case "$dest" in /*) ;; *) dest="$(dirname "$link")/$dest" ;; esac
d="${dest%/*}"; b="${dest##*/}"
if d_real="$(cd "$d" 2>/dev/null && pwd -P)"; then printf '%s\n' "$d_real/$b"; else printf '%s\n' "$dest"; fi
}
_gstack_target_is_ours() {
# $1 = absolute target path, $2 = gstack payload dir
local t="$1" g="$2" g_real render render_real
g_real="$(cd "$g" 2>/dev/null && pwd -P || printf '%s' "$g")"
render="${GSTACK_USER_RENDER_DIR:-${GSTACK_HOME:-$HOME/.gstack}/render/claude}"
render_real="$(cd "$render" 2>/dev/null && pwd -P || printf '%s' "$render")"
case "$t" in
"$g"/*|"$g_real"/*|"$render"/*|"$render_real"/*|gstack/*|*/gstack/*|*/.gstack/render/claude/*) return 0 ;;
esac
# A checkout named without a `gstack` segment (git worktree add
# ../gstack-<branch>, a ZIP unpacked as gstack-main): the target's skill
# root is a gstack tree if it carries setup + VERSION + bin/gstack-relink
# (a hand-written skill repo with a VERSION file and a setup script does not).
local root="${t%/*/SKILL.md}"
if [ "$root" != "$t" ] && [ -f "$root/VERSION" ] && [ -f "$root/setup" ] && [ -f "$root/bin/gstack-relink" ]; then return 0; fi
return 1
}
_claude_entry_is_ours() {
# $1 = existing entry (dir or symlink), $2 = the gstack source SKILL.md it
# would be linked to, $3 = gstack payload dir
local entry="$1" src_md="$2" g="$3" render_md
_claude_entry_owned_strongly "$entry" "$g" && return 0
# A symlink that did not resolve into gstack is someone else's; never follow
# it into the "unclaimed directory" rule below.
[ -L "$entry" ] && return 1
# No SKILL.md at all: an UNCLAIMED directory (a weak cleanup left the user's
# other files behind, or it was never a skill). Adding our SKILL.md
# overwrites nothing, so installing into it is allowed; the cleanup arms
# require a SKILL.md and so never touch it.
if [ -d "$entry" ] && [ ! -e "$entry/SKILL.md" ] && [ ! -L "$entry/SKILL.md" ]; then return 0; fi
if [ -d "$entry" ] && [ -f "$entry/SKILL.md" ] && [ ! -L "$entry/SKILL.md" ]; then
[ -n "$src_md" ] && [ -f "$src_md" ] && cmp -s "$entry/SKILL.md" "$src_md" && return 0
# A gbrain install serves the RENDERED file (link_claude_skill_dirs prefers
# it), so an exact copy of that render is ours too.
if [ -n "$src_md" ]; then
render_md="${GSTACK_USER_RENDER_DIR:-${GSTACK_HOME:-$HOME/.gstack}/render/claude}/$(basename "$(dirname "$src_md")")/SKILL.md"
[ -f "$render_md" ] && cmp -s "$entry/SKILL.md" "$render_md" && return 0
fi
_gstack_generated_header "$entry/SKILL.md" && return 0
fi
return 1
}
# _claude_entry_owned_strongly ENTRY GSTACK_DIR — we created it: a symlink into
# gstack, or a real dir with the .gstack-owned marker or a SKILL.md symlink
# into gstack. Only strong proof authorizes deleting a directory whole.
_claude_entry_owned_strongly() {
local entry="$1" g="$2" dest
if [ -L "$entry" ]; then
dest="$(_gstack_link_target_abs "$entry")" || return 1
_gstack_target_is_ours "$dest" "$g"; return $?
fi
[ -d "$entry" ] || return 1
[ -f "$entry/.gstack-owned" ] && return 0
if [ -L "$entry/SKILL.md" ]; then
dest="$(_gstack_link_target_abs "$entry/SKILL.md")" || return 1
_gstack_target_is_ours "$dest" "$g"; return $?
fi
return 1
}
# Weakly-proven real files we would otherwise overwrite are moved here (mv, so
# the path is free for the link); the final summary prints one line.
_SKILL_BACKUP_ROOT="${GSTACK_HOME:-$HOME/.gstack}/backups/skills/$(date +%Y%m%dT%H%M%S)"
_BACKED_UP_SKILL_MDS=()
_backup_skill_md() {
# Returns non-zero when the file could NOT be moved: the caller must then
# leave the entry untouched (a failed backup is never a license to overwrite).
local file="$1" name="$2"
mkdir -p "$_SKILL_BACKUP_ROOT/$name" 2>/dev/null || return 1
mv -f "$file" "$_SKILL_BACKUP_ROOT/$name/SKILL.md" 2>/dev/null || return 1
_BACKED_UP_SKILL_MDS+=("$name")
return 0
}
# _cleanup_weak_dir DIR — remove only what weak proof covers: the SKILL.md and
# our marker. User files in the directory stay, and so does the directory
# when it is not empty afterwards.
# _cleanup_weak_dir DIR GSTACK_DIR [SRC_SKILL_MD NAME] — remove only what weak
# proof covers. A real SKILL.md that differs from our source (raw, or with its
# name: line rewritten to NAME, which is how alias and prefixed copies differ)
# is a customized file: it is moved to the backup root, never deleted, and if
# the backup fails it stays. Our runtime-asset links go; the user's files stay.
_cleanup_weak_dir() {
local d="$1" g="$2" src="${3:-}" name="${4:-${1##*/}}" e dest
if [ -f "$d/SKILL.md" ] && [ ! -L "$d/SKILL.md" ] && [ -n "$src" ] && [ -f "$src" ] \
&& ! cmp -s "$d/SKILL.md" "$src" \
&& ! sed "1,/^---\$/ s/^name:[[:space:]].*/name: $name/" "$src" | cmp -s - "$d/SKILL.md"; then
if ! _backup_skill_md "$d/SKILL.md" "$name"; then
echo " kept $name/SKILL.md: could not back up the customized file — left untouched" >&2
return 0
fi
else
rm -f "$d/SKILL.md"
fi
rm -f "$d/.gstack-owned"
for e in "$d"/* "$d"/.[!.]* "$d"/..?*; do
[ -L "$e" ] || continue
dest="$(_gstack_link_target_abs "$e")" || continue
if _gstack_target_is_ours "$dest" "$g"; then rm -f "$e"; fi
done
rmdir "$d" 2>/dev/null || echo " cleaned ${d##*/}/SKILL.md (other files in that directory were left in place)"
}
# _gstack_dir_only_links DIR GSTACK_DIR — true when deleting DIR whole loses
# nothing of the user's: every entry is a symlink resolving into gstack, or
# our marker. A user's own link (notes.md -> ~/notes) makes the dir mixed.
_gstack_dir_only_links() {
local d="$1" g="$2" e dest
for e in "$d"/* "$d"/.[!.]* "$d"/..?*; do
{ [ -e "$e" ] || [ -L "$e" ]; } || continue
[ "${e##*/}" = ".gstack-owned" ] && continue
[ -L "$e" ] || return 1
dest="$(_gstack_link_target_abs "$e")" || return 1
_gstack_target_is_ours "$dest" "$g" || return 1
done
return 0
}
# _cleanup_linked_dir DIR GSTACK_DIR — a real dir whose SKILL.md is a symlink
# into gstack. Whole-directory removal needs the marker (we created it) or a
# directory holding nothing but our links; otherwise only our files go.
_cleanup_linked_dir() {
if [ -f "$1/.gstack-owned" ] || _gstack_dir_only_links "$1" "$2"; then rm -rf "$1"; else _cleanup_weak_dir "$1" "$2"; fi
}
# _gstack_generated_header FILE — a pre-marker legacy COPY (Windows, before
# .gstack-owned existed) is recognized by gen-skill-docs' full two-line banner
# near the top, not by a one-line substring another generator could plausibly
# emit. Still forgeable by a gstack fork that renders the same banner — that
# residual is accepted and filed; the marker is the load-bearing signal.
_gstack_generated_header() {
# Bytes, not lines: a long frontmatter pushes the banner past line 40 in
# four real skills (investigate: line 57), and a line-count check left them
# "foreign" on every pre-marker Windows install.
local f="$1" head40
head40="$(head -c 8192 "$f" 2>/dev/null)" || return 1
case "$head40" in
*'<!-- AUTO-GENERATED from '*'<!-- Regenerate: bun run gen:skill-docs -->'*) return 0 ;;
esac
return 1
}
_write_owned_marker() {
# Windows copy installs have no symlink to readlink; the marker proves
# provenance. Records the owning payload's real path for forensics.
local dir="$1" g="$2"
printf '%s\n' "$(cd "$g" 2>/dev/null && pwd -P || printf '%s' "$g")" > "$dir/.gstack-owned" 2>/dev/null || true
}
# ─── Ownership gates for the Windows refresh bypass (#2444 → #2142) ─────────
# On Windows a refresh means rm -rf + re-copy (_link_or_copy). The host
# skills dirs are SHARED namespaces (~/.codex/skills, ~/.factory/skills,
@@ -420,6 +591,13 @@ _kill_tree() {
for child in $(pgrep -P "$pid" 2>/dev/null); do
_kill_tree "$child"
done
elif [ -d /proc ]; then
# debian-slim and git-bash ship no pgrep: walk /proc for children. The
# comm field "(name)" may contain spaces and parens, so strip through the
# LAST closing paren (proc(5)) before reading the ppid (second field after it).
for child in $(awk -v p="$pid" '{ s=$0; sub(/^.*\) /, "", s); split(s, f, " "); if (f[2]==p) print $1 }' /proc/[0-9]*/stat 2>/dev/null); do
_kill_tree "$child"
done
fi
kill -9 "$pid" 2>/dev/null || true
}
@@ -776,7 +954,42 @@ if [ -f /etc/os-release ]; then
fi
fi
if ! ensure_playwright_browser; then
# Chromium is BEST-EFFORT (#1900, #1901, #1902, #913, #2233). Every later step
# — skill registration (# 4), Codex/Kiro installs, migrations, hooks — is
# independent of the browser, so a failed or wedged download must never abort
# setup under `set -e`. Each failure records a reason code in _PW_FAIL_REASON;
# the skills that need Chromium (/qa, /qa-only, /design-review, /browse,
# make-pdf, /pair-agent) are named in the final summary instead of the user
# discovering a half-installed gstack. Lock contention is a reason too: another
# setup is installing Chromium right now, so this run registers skills and
# re-probes next time. The download is bounded (default 600s, env
# GSTACK_PLAYWRIGHT_INSTALL_TIMEOUT) because Playwright's own retries cover a
# flaky socket but not a wedged bunx; a wedged installer is killed with its
# child tree (_kill_tree: pgrep-walked, /proc-walked where pgrep is missing).
# Reason codes: skipped, chromium-install,
# chromium-install-timeout, chromium-install-locked, windows-no-node,
# windows-node-modules, post-install-launch.
# test/setup-playwright-best-effort.test.ts pins this block.
_PW_FAIL_REASON=""
_pw_fail() {
local code="$1"; shift
_PW_FAIL_REASON="${_PW_FAIL_REASON:+$_PW_FAIL_REASON,}$code"
echo " Chromium bootstrap: $code — $*" >&2
}
_PW_INSTALL_TIMEOUT="${GSTACK_PLAYWRIGHT_INSTALL_TIMEOUT:-600}"
# Normalize to a plain positive integer or fall back to the default: empty and
# non-numeric are garbage; "0"/"000" would kill the install on the first poll;
# "0600" is 600; anything past nine digits is not a deadline (a value bash
# cannot compare would leave the install unbounded — the exact failure the
# bound exists to prevent).
case "$_PW_INSTALL_TIMEOUT" in ''|*[!0-9]*) _PW_INSTALL_TIMEOUT=600 ;; esac
[ "${#_PW_INSTALL_TIMEOUT}" -le 9 ] || _PW_INSTALL_TIMEOUT=600
_PW_INSTALL_TIMEOUT=$((10#$_PW_INSTALL_TIMEOUT))
[ "$_PW_INSTALL_TIMEOUT" -gt 0 ] || _PW_INSTALL_TIMEOUT=600
if [ "${GSTACK_SKIP_PLAYWRIGHT:-0}" = "1" ]; then
_pw_fail skipped "GSTACK_SKIP_PLAYWRIGHT=1 — Chromium install skipped by request (#913)"
elif ! ensure_playwright_browser; then
echo "Installing Playwright Chromium..."
# XProtect self-heal (#2554): the probe failure may be the OS killing the
# cached Chromium, not a missing install. Clear quarantine on the Playwright
@@ -786,11 +999,47 @@ if ! ensure_playwright_browser; then
# Stale-lock self-heal: a SIGKILL'd prior setup leaves the lock dir behind
# forever (mkdir mutexes have no owner). If the recorded holder PID is dead,
# reclaim instead of telling the user to rmdir by hand.
if [ -d "$_PW_LOCK" ] && [ -f "$_PW_LOCK/pid" ]; then
_PW_HOLDER=$(cat "$_PW_LOCK/pid" 2>/dev/null || true)
if [ -n "$_PW_HOLDER" ] && ! kill -0 "$_PW_HOLDER" 2>/dev/null; then
echo " reclaiming stale Chromium-install lock (holder pid $_PW_HOLDER is gone)" >&2
rm -rf "$_PW_LOCK" 2>/dev/null || true
if [ -d "$_PW_LOCK" ]; then
_PW_STALE=0; _PW_HOLDER=0
_PW_LOCK_OLD=""
[ -n "$(find "$_PW_LOCK" -maxdepth 0 -mmin +$(( _PW_INSTALL_TIMEOUT / 60 + 1 )) 2>/dev/null)" ] && _PW_LOCK_OLD=1
if [ ! -f "$_PW_LOCK/pid" ]; then
# No holder recorded (killed between mkdir and echo, or mid-write by a
# live setup): nothing to probe, so only age can prove abandonment.
[ -n "$_PW_LOCK_OLD" ] && _PW_STALE=1
else
_PW_HOLDER=$(cat "$_PW_LOCK/pid" 2>/dev/null || true)
# A pid must be a positive integer: "", "-1" (kill -0 -1 signals every
# process and "succeeds") or "0" (the process group) are stale, not live.
case "$_PW_HOLDER" in ''|*[!0-9]*) _PW_HOLDER=0 ;; esac
if [ "$_PW_HOLDER" -eq 0 ] || ! kill -0 "$_PW_HOLDER" 2>/dev/null; then
_PW_STALE=1
elif [ -n "$_PW_LOCK_OLD" ]; then
# The holder is alive but the lock is older than the install bound: the
# holder is past its own deadline, or its pid was recycled to an
# unrelated long-lived process. Either way nobody is installing.
_PW_STALE=1
fi
fi
if [ "$_PW_STALE" -eq 1 ]; then
echo " reclaiming stale Chromium-install lock (holder pid ${_PW_HOLDER:-?} is gone or past the install bound)" >&2
# Rename first: two setups judging the same lock stale race on rm -rf +
# mkdir, and the loser would delete the winner's fresh lock. mv of a
# directory is atomic, so exactly one of them reclaims — and if the dir
# we moved already belongs to a NEW live holder (it re-created the lock
# between our judgment and our mv), hand it straight back.
if mv "$_PW_LOCK" "$_PW_LOCK.stale.$$" 2>/dev/null; then
_PW_MOVED_PID=$(cat "$_PW_LOCK.stale.$$/pid" 2>/dev/null || true)
case "$_PW_MOVED_PID" in ''|*[!0-9]*) _PW_MOVED_PID=0 ;; esac
if { [ "$_PW_MOVED_PID" -gt 0 ] && [ "$_PW_MOVED_PID" != "$_PW_HOLDER" ] && kill -0 "$_PW_MOVED_PID" 2>/dev/null; } \
|| { [ ! -f "$_PW_LOCK.stale.$$/pid" ] && [ -z "$_PW_LOCK_OLD" ]; }; then
# A new live holder, or a fresh lock whose holder has not written its
# pid yet (mkdir done, echo pending): not ours to reclaim.
mv "$_PW_LOCK.stale.$$" "$_PW_LOCK" 2>/dev/null || true
else
rm -rf "$_PW_LOCK.stale.$$" 2>/dev/null || true
fi
fi
fi
fi
if mkdir "$_PW_LOCK" 2>/dev/null; then
@@ -806,48 +1055,58 @@ if ! ensure_playwright_browser; then
else
bunx playwright install chromium
fi
)
) &
_PW_PID=$!
# Ctrl-C during the download: a backgrounded child ignores SIGINT, so
# without this the installer would keep running as an orphan while the
# EXIT trap frees the lock — the #2136 pile-up the lock exists to prevent.
trap '_kill_tree "$_PW_PID" 2>/dev/null; rm -rf "$_PW_LOCK" 2>/dev/null || true; cleanup_copied_bun; exit 130' INT TERM
_PW_RC=0
_wait_with_deadline "$_PW_PID" "$_PW_INSTALL_TIMEOUT" || _PW_RC=$?
trap - INT TERM
if [ "$_PW_RC" -eq 124 ]; then
_pw_fail chromium-install-timeout "bunx playwright install chromium exceeded ${_PW_INSTALL_TIMEOUT}s and was killed (raise with GSTACK_PLAYWRIGHT_INSTALL_TIMEOUT=<seconds>)"
elif [ "$_PW_RC" -ne 0 ]; then
_pw_fail chromium-install "bunx playwright install chromium exited $_PW_RC (offline, proxy, or blocked download?)"
fi
rm -rf "$_PW_LOCK" 2>/dev/null || true
# Restore the original handler (never `trap - EXIT`, which would clear
# cleanup_copied_bun for the rest of the script).
trap cleanup_copied_bun EXIT
else
echo " another gstack setup is already installing Chromium (lock: $_PW_LOCK)." >&2
echo " Wait for it to finish, then re-run ./setup. If no other setup is running," >&2
echo " remove the stale lock: rm -rf \"$_PW_LOCK\"" >&2
exit 1
_pw_fail chromium-install-locked "another gstack setup is installing Chromium (lock: $_PW_LOCK) — re-run ./setup after it finishes, or remove a stale lock: rm -rf \"$_PW_LOCK\""
fi
if [ "$IS_WINDOWS" -eq 1 ]; then
if [ -z "$_PW_FAIL_REASON" ] && [ "$IS_WINDOWS" -eq 1 ]; then
# On Windows, Node.js launches Chromium (not Bun — see oven-sh/bun#4253).
# Ensure playwright is importable by Node from the gstack directory.
if ! command -v node >/dev/null 2>&1; then
echo "gstack setup failed: Node.js is required on Windows (Bun cannot launch Chromium due to a pipe bug)" >&2
echo " Install Node.js: https://nodejs.org/" >&2
exit 1
_pw_fail windows-no-node "Node.js is required on Windows to launch Chromium (Bun cannot: oven-sh/bun#4253) — install from https://nodejs.org/ and re-run ./setup"
else
echo "Windows detected — verifying Node.js can load Playwright..."
if ! (
cd "$SOURCE_GSTACK_DIR"
# Bun's node_modules already has playwright; verify Node can require it.
# @ngrok/ngrok is externalized in server-node.mjs and resolved at runtime;
# verify the platform-specific native binary is installed so /pair-agent
# tunnels don't fail later with a cryptic module-not-found error.
# &&-chained: errexit is off inside an `if` condition, so a failed npm
# install on the first line must not be masked by the second.
{ node -e "require('playwright')" 2>/dev/null || npm install --no-save playwright; } &&
{ node -e "require('@ngrok/ngrok')" 2>/dev/null || npm install --no-save @ngrok/ngrok; }
); then
_pw_fail windows-node-modules "npm could not install playwright / @ngrok/ngrok for Node.js"
fi
fi
echo "Windows detected — verifying Node.js can load Playwright..."
(
cd "$SOURCE_GSTACK_DIR"
# Bun's node_modules already has playwright; verify Node can require it
node -e "require('playwright')" 2>/dev/null || npm install --no-save playwright
# @ngrok/ngrok is externalized in server-node.mjs and resolved at runtime.
# Verify the platform-specific native binary is installed so /pair-agent
# tunnels don't fail later with a cryptic module-not-found error.
node -e "require('@ngrok/ngrok')" 2>/dev/null || npm install --no-save @ngrok/ngrok
)
fi
fi
if ! ensure_playwright_browser; then
if [ -z "$_PW_FAIL_REASON" ] && ! ensure_playwright_browser; then
if [ "$IS_WINDOWS" -eq 1 ]; then
echo "gstack setup failed: Playwright Chromium could not be launched via Node.js" >&2
echo " This is a known issue with Bun on Windows (oven-sh/bun#4253)." >&2
echo " Ensure Node.js is installed and 'node -e \"require('playwright')\"' works." >&2
_pw_fail post-install-launch "Playwright Chromium could not be launched via Node.js (oven-sh/bun#4253) — ensure 'node -e \"require('playwright')\"' works, then re-run ./setup"
else
echo "gstack setup failed: Playwright Chromium could not be launched" >&2
_pw_fail post-install-launch "Playwright Chromium installed but could not be launched — on Ubuntu 24.04+ (AppArmor blocks unprivileged user namespaces) try GSTACK_CHROMIUM_NO_SANDBOX=1 (#2157)"
fi
exit 1
fi
# 2b. Ensure a color-emoji font is installed so make-pdf emoji render (Linux).
@@ -859,7 +1118,9 @@ if ! ensure_emoji_font; then
echo " Fedora: sudo dnf install google-noto-color-emoji-fonts" >&2
echo " Arch: sudo pacman -S noto-fonts-emoji" >&2
echo " Alpine: sudo apk add font-noto-emoji" >&2
else
elif [ -z "$_PW_FAIL_REASON" ]; then
# Only when Chromium is actually usable — restarting a daemon that cannot
# launch its browser just produces a second failure line.
refresh_browse_daemon_for_fonts
fi
@@ -885,6 +1146,12 @@ mkdir -p "$HOME/.gstack/projects"
_link_skill_runtime_assets() {
local src_dir="$1"
local dst_dir="$2"
# $3 = 0 when the destination directory is not provably ours (pre-existed
# unclaimed, or only weakly proven where gstack never wrote real assets): its
# real files are the user's, so a same-named real asset is kept and reported
# instead of replaced (#2119). Symlinks are never content and are always
# refreshed. Default 1 = the directory is ours.
local replace_real="${3:-1}"
local asset asset_name
for asset in "$src_dir"/*; do
[ -e "$asset" ] || continue # empty-glob guard
@@ -892,8 +1159,12 @@ _link_skill_runtime_assets() {
case "$asset_name" in
SKILL.md|node_modules|dist|test|*.tmpl) continue ;;
esac
# Refresh unconditionally: rm the old entry (symlink OR real copy — the
# Windows install pattern) so re-runs after `git pull` pick up changes.
if [ -e "$dst_dir/$asset_name" ] && [ ! -L "$dst_dir/$asset_name" ] && [ "$replace_real" != "1" ]; then
echo " kept ${dst_dir##*/}/$asset_name: a file you own already uses that name — left untouched" >&2
continue
fi
# Refresh: rm the old entry (symlink OR real copy — the Windows install
# pattern) so re-runs after `git pull` pick up changes.
if [ -e "$dst_dir/$asset_name" ] || [ -L "$dst_dir/$asset_name" ]; then
rm -rf "$dst_dir/$asset_name"
fi
@@ -917,6 +1188,27 @@ _link_skill_runtime_assets() {
# gstack/ (which would auto-prefix them as gstack-*).
# When SKILL_PREFIX=1, directories are prefixed with "gstack-".
# Use --no-prefix to restore flat names.
# Run gstack-relink and surface only what the user must see: foreign entries it
# skipped (deduped against the ones this setup already reported, same wording)
# and any pre-existing SKILL.md it moved to the backup root.
_run_relink_quiet() {
local out line name
out="$(GSTACK_SKILLS_DIR="$INSTALL_SKILLS_DIR" GSTACK_INSTALL_DIR="$SOURCE_GSTACK_DIR" "$GSTACK_RELINK" 2>&1 || true)"
while IFS= read -r line; do
case "$line" in
' skipped '*)
name="${line# skipped }"; name="${name%%:*}"
case " ${_FOREIGN_SKIPPED_ENTRIES[*]:-} " in
*" $name "*) ;;
*) echo "$line" >&2; _FOREIGN_SKIPPED_ENTRIES+=("$name") ;;
esac ;;
'Moved '*|' cleaned '*) echo " ${line# }" >&2 ;;
esac
done <<EOF
$out
EOF
}
link_claude_skill_dirs() {
local gstack_dir="$1"
local skills_dir="$2"
@@ -939,6 +1231,32 @@ link_claude_skill_dirs() {
link_name="$skill_name"
fi
target="$skills_dir/$link_name"
# #2119: a destination that exists and is NOT ours is a user's skill that
# shares our name. Never rm/mkdir/ln into it — skip and report.
if { [ -e "$target" ] || [ -L "$target" ]; } && ! _claude_entry_is_ours "$target" "$gstack_dir/$dir_name/SKILL.md" "$gstack_dir"; then
echo " skipped $link_name: existing entry is not gstack-managed (foreign skill with the same name) — left untouched" >&2
_FOREIGN_SKIPPED_ENTRIES+=("$link_name")
continue
fi
# Remember whether WE are creating this directory: only then may the
# provenance marker below make it deletable whole. A directory we merely
# link into (unclaimed, or a legacy install) never gets one — legacy
# all-links dirs are removed by the only-links rule instead.
# _assets_replace: may _link_skill_runtime_assets replace a REAL file or
# dir already present under the target? Yes for a directory we create or
# strongly own (marker, or SKILL.md symlink into gstack). On Windows also
# for a weakly-proven real-file copy install (the legacy pre-marker shape,
# whose asset copies are ours). Otherwise (unclaimed, or a weak copy on a
# platform where gstack never wrote real assets) real files are the
# user's and are kept.
_pre_exists=0; _assets_replace=1
if [ -e "$target" ] || [ -L "$target" ]; then
_pre_exists=1
if _claude_entry_owned_strongly "$target" "$gstack_dir"; then _assets_replace=1
elif [ "$IS_WINDOWS" -eq 1 ] && [ -f "$target/SKILL.md" ] && [ ! -L "$target/SKILL.md" ]; then _assets_replace=1
else _assets_replace=0
fi
fi
# Upgrade old directory symlinks to real directories
if [ -L "$target" ]; then
rm -f "$target"
@@ -959,7 +1277,23 @@ link_claude_skill_dirs() {
if [ -f "$_render_dir/$dir_name/SKILL.md" ]; then
_skill_md_src="$_render_dir/$dir_name/SKILL.md"
fi
# A real-file SKILL.md we can only WEAKLY prove ours and whose content
# differs from what we are about to serve is moved aside, not overwritten.
if [ -f "$target/SKILL.md" ] && [ ! -L "$target/SKILL.md" ] && ! _claude_entry_owned_strongly "$target" "$gstack_dir" \
&& ! cmp -s "$target/SKILL.md" "$_skill_md_src"; then
if ! _backup_skill_md "$target/SKILL.md" "$link_name"; then
echo " skipped $link_name: could not back up its customized SKILL.md — left untouched" >&2
_FOREIGN_SKIPPED_ENTRIES+=("$link_name")
continue
fi
fi
_link_or_copy "$_skill_md_src" "$target/SKILL.md"
# Provenance marker (#2119) on every platform — path-independent proof
# for Windows copies and for checkouts whose path carries no `gstack`
# segment — but only for a directory we created or already owned: a
# pre-existing unclaimed or weakly-proven directory must not become
# deletable whole because we linked one file into it.
if [ "$_pre_exists" -eq 0 ] || [ -f "$target/.gstack-owned" ]; then _write_owned_marker "$target" "$gstack_dir"; fi
# Link every runtime asset the skill ships next to its SKILL.md (#2317,
# #2454): sections/ for carved skills, review's checklist.md +
# specialists/, qa's templates/ + references/, gstack-upgrade's
@@ -967,7 +1301,7 @@ link_claude_skill_dirs() {
# landed and /review 404'd at "Read .claude/skills/review/checklist.md"
# on every fresh Claude install. Routes through _link_or_copy so Windows
# gets real copies refreshed on every ./setup.
_link_skill_runtime_assets "$gstack_dir/$dir_name" "$target"
_link_skill_runtime_assets "$gstack_dir/$dir_name" "$target" "$_assets_replace"
linked+=("$link_name")
fi
done
@@ -993,13 +1327,27 @@ _install_alias_skill_md() {
local dst_dir="$2"
local alias_name="$3"
[ -f "$src_skill_md" ] || return 0
# #2119: an existing alias-named entry that is not ours is a user's skill.
# (Ours: a whole-dir symlink into gstack, or a copy carrying the generated
# header — every alias copy does.)
if { [ -e "$dst_dir" ] || [ -L "$dst_dir" ]; } && ! _claude_entry_is_ours "$dst_dir" "$src_skill_md" "$SOURCE_GSTACK_DIR"; then
echo " skipped $alias_name: existing entry is not gstack-managed (foreign skill with the same name) — left untouched" >&2
_FOREIGN_SKIPPED_ENTRIES+=("$alias_name")
return 0
fi
# Old installs left the alias as a whole-dir symlink — replace it.
_alias_pre=0
if [ -e "$dst_dir" ] || [ -L "$dst_dir" ]; then _alias_pre=1; fi
if [ -L "$dst_dir" ]; then rm -f "$dst_dir"; fi
mkdir -p "$dst_dir"
# Remove any prior symlinked SKILL.md so the redirect below cannot write
# through it into the generated source.
rm -f "$dst_dir/SKILL.md"
sed "1,/^---\$/ s/^name:[[:space:]].*/name: $alias_name/" "$src_skill_md" > "$dst_dir/SKILL.md"
# A rewritten copy is a real file on every platform; the marker proves it
# ours on the next run without leaning on the banner — but only for a
# directory we created (or already marked), never one we merely wrote into.
if [ "$_alias_pre" -eq 0 ] || [ -f "$dst_dir/.gstack-owned" ]; then _write_owned_marker "$dst_dir" "$SOURCE_GSTACK_DIR"; fi
}
# Claude Code skips the repo-shaped ~/.claude/skills/gstack directory when
@@ -1055,17 +1403,20 @@ cleanup_old_claude_symlinks() {
# render prefix (~/.gstack/render/claude/...), which is not `/gstack/`.
case "$link_dest" in
gstack/*|*/gstack/*|*/.gstack/render/claude/*)
rm -rf "$old_target"
_cleanup_linked_dir "$old_target" "$gstack_dir"
removed+=("$skill_name")
;;
esac
fi
done
# Windows install pattern: real dir with real-file SKILL.md (no symlink
# available, so we can't readlink to verify provenance). Iterate known
# gstack skill names from "$gstack_dir"/*, so a name match plus IS_WINDOWS
# is safe to treat as gstack-managed during a mode flip. When the payload
# is gone this branch is a no-op — a real file has no proven owner.
# available, so we can't readlink to verify provenance). A bare name match
# deleted a user's own same-name skill (#2119); ownership is now proven by
# the .gstack-owned marker link_claude_skill_dirs writes, or — for copies
# made before the marker existed — by the copy being byte-identical to the
# gstack source SKILL.md or carrying gen-skill-docs' AUTO-GENERATED header
# (every generated SKILL.md does; a hand-written skill does not). Anything
# else is foreign and is left alone.
if [ "${IS_WINDOWS:-0}" -eq 1 ] && [ -d "$gstack_dir" ]; then
for skill_dir in "$gstack_dir"/*/; do
if [ -f "$skill_dir/SKILL.md" ]; then
@@ -1074,8 +1425,13 @@ cleanup_old_claude_symlinks() {
case "$skill_name" in gstack-*) continue ;; esac
old_target="$skills_dir/$skill_name"
if [ -d "$old_target" ] && [ ! -L "$old_target" ] \
&& [ -f "$old_target/SKILL.md" ] && [ ! -L "$old_target/SKILL.md" ]; then
rm -rf "$old_target"
&& [ -f "$old_target/SKILL.md" ] && [ ! -L "$old_target/SKILL.md" ] \
&& { [ -f "$old_target/.gstack-owned" ] \
|| cmp -s "$old_target/SKILL.md" "$skill_dir/SKILL.md" \
|| _gstack_generated_header "$old_target/SKILL.md"; }; then
# Only the marker proves we created the directory; weak proof covers
# the SKILL.md alone (a user's files next to it survive).
if [ -f "$old_target/.gstack-owned" ]; then rm -rf "$old_target"; else _cleanup_weak_dir "$old_target" "$gstack_dir" "$skill_dir/SKILL.md" "$skill_name"; fi
removed+=("$skill_name")
fi
fi
@@ -1101,11 +1457,13 @@ cleanup_prefixed_claude_symlinks() {
# (e.g., remove gstack-qa but NOT gstack-upgrade which is the real dir name)
case "$skill_name" in gstack-*) continue ;; esac
prefixed_target="$skills_dir/gstack-$skill_name"
# Remove directory symlinks pointing into gstack/
# Remove directory symlinks pointing into gstack/ — anchored path
# segments, same as cleanup_old_claude_symlinks and gstack-uninstall: a
# bare *gstack* substring would wipe a user skill under ~/tools/gstack-fork/.
if [ -L "$prefixed_target" ]; then
link_dest="$(readlink "$prefixed_target" 2>/dev/null || true)"
case "$link_dest" in
gstack/*|*/gstack/*)
gstack/*|*/gstack/*|*/.gstack/render/claude/*)
rm -f "$prefixed_target"
removed+=("gstack-$skill_name")
;;
@@ -1114,16 +1472,20 @@ cleanup_prefixed_claude_symlinks() {
elif [ -d "$prefixed_target" ] && [ -L "$prefixed_target/SKILL.md" ]; then
link_dest="$(readlink "$prefixed_target/SKILL.md" 2>/dev/null || true)"
case "$link_dest" in
*gstack*)
rm -rf "$prefixed_target"
gstack/*|*/gstack/*|*/.gstack/render/claude/*)
_cleanup_linked_dir "$prefixed_target" "$gstack_dir"
removed+=("gstack-$skill_name")
;;
esac
# Windows install pattern: real dir with real-file SKILL.md. Same
# reasoning as cleanup_old_claude_symlinks — directory name match plus
# IS_WINDOWS is safe during a mode flip.
elif [ "$IS_WINDOWS" -eq 1 ] && [ -d "$prefixed_target" ] && [ -f "$prefixed_target/SKILL.md" ]; then
rm -rf "$prefixed_target"
# Windows install pattern: real dir with real-file SKILL.md. Provenance
# must be PROVEN (#2119), never assumed from the name: the marker
# link_claude_skill_dirs writes, a byte-identical copy of the source, or
# gen-skill-docs' generated header (legacy copies made before the marker).
elif [ "$IS_WINDOWS" -eq 1 ] && [ -d "$prefixed_target" ] && [ -f "$prefixed_target/SKILL.md" ] && [ ! -L "$prefixed_target/SKILL.md" ] \
&& { [ -f "$prefixed_target/.gstack-owned" ] \
|| cmp -s "$prefixed_target/SKILL.md" "$skill_dir/SKILL.md" \
|| _gstack_generated_header "$prefixed_target/SKILL.md"; }; then
if [ -f "$prefixed_target/.gstack-owned" ]; then rm -rf "$prefixed_target"; else _cleanup_weak_dir "$prefixed_target" "$gstack_dir" "$skill_dir/SKILL.md" "gstack-$skill_name"; fi
removed+=("gstack-$skill_name")
fi
fi
@@ -1681,7 +2043,7 @@ if [ "$INSTALL_CLAUDE" -eq 1 ]; then
# setup, stale git state, or gen:skill-docs left name: fields out of sync.
GSTACK_RELINK="$SOURCE_GSTACK_DIR/bin/gstack-relink"
if [ -x "$GSTACK_RELINK" ]; then
GSTACK_SKILLS_DIR="$INSTALL_SKILLS_DIR" GSTACK_INSTALL_DIR="$SOURCE_GSTACK_DIR" "$GSTACK_RELINK" >/dev/null 2>&1 || true
_run_relink_quiet
fi
# Backwards-compat alias: /connect-chrome → /open-gstack-browser
# Rewritten copy, not a symlink: a symlinked alias re-serves the canonical
@@ -1754,7 +2116,7 @@ if [ "$INSTALL_CLAUDE" -eq 1 ]; then
_CLAUDE_SKILLS_LINKED=1
GSTACK_RELINK="$SOURCE_GSTACK_DIR/bin/gstack-relink"
if [ -x "$GSTACK_RELINK" ]; then
GSTACK_SKILLS_DIR="$INSTALL_SKILLS_DIR" GSTACK_INSTALL_DIR="$SOURCE_GSTACK_DIR" "$GSTACK_RELINK" >/dev/null 2>&1 || true
_run_relink_quiet
fi
# Rewritten copy, not a symlink: a symlinked alias re-serves the
# canonical name: open-gstack-browser, so one of the two silently
@@ -1981,7 +2343,7 @@ if [ ! -f "$HOME/.gstack/.welcome-seen" ]; then
log ""
# Best-effort onboarding telemetry (respects telemetry!=off; never blocks setup).
if [ -x "$SOURCE_GSTACK_DIR/bin/gstack-telemetry-log" ]; then
"$SOURCE_GSTACK_DIR/bin/gstack-telemetry-log" --event-type onboarding --skill _setup_welcome --outcome shown >/dev/null 2>&1 || true
"$SOURCE_GSTACK_DIR/bin/gstack-telemetry-log" --event-type onboarding --skill _setup_welcome --outcome shown --no-sweep >/dev/null 2>&1 || true
fi
touch "$HOME/.gstack/.welcome-seen"
fi
@@ -2587,3 +2949,41 @@ if ! grep -q '^redact_prepush_hook:' "$_GSTACK_CFG_FILE" 2>/dev/null; then
log " installs the hook automatically in every repo you ship from."
fi
fi
# ─── Chromium bootstrap summary (best-effort browser, see # 2) ───────────────
# Printed LAST so it is the thing the user sees, after every skill registered.
_PW_BROWSER_SKILLS="/qa, /qa-only, /design-review, /browse, make-pdf, /pair-agent, and any other skill that drives the browser"
if [ "${_PW_FAIL_REASON:-}" = "skipped" ]; then
# An explicit opt-out is not a failure: say what is unavailable and stop.
log ""
log "Chromium install skipped by request (GSTACK_SKIP_PLAYWRIGHT=1)."
log " Browser skills ($_PW_BROWSER_SKILLS) need it; re-run ./setup without the flag when you want them."
elif [ -n "${_PW_FAIL_REASON:-}" ]; then
log ""
log "Browser unavailable: Chromium bootstrap did not complete ($_PW_FAIL_REASON)."
log " Skills that need it: $_PW_BROWSER_SKILLS."
log " Everything else is installed and works. Fix the cause and re-run ./setup."
case "$_PW_FAIL_REASON" in
*chromium-install-timeout*) log " Slow link? Raise the bound: GSTACK_PLAYWRIGHT_INSTALL_TIMEOUT=1800 ./setup" ;;
esac
case "$_PW_FAIL_REASON" in
*post-install-launch*) log " Ubuntu 24.04+ (AppArmor user namespaces): GSTACK_CHROMIUM_NO_SANDBOX=1 ./setup (#2157)" ;;
esac
# Reason code only — never a path, hostname, or the installer's output. The
# event is a one-shot with no session of its own, so --no-sweep keeps it
# from finalizing other live sessions' in-flight .pending markers as
# outcome:unknown. Telemetry-gated inside gstack-telemetry-log.
if [ -x "$SOURCE_GSTACK_DIR/bin/gstack-telemetry-log" ]; then
"$SOURCE_GSTACK_DIR/bin/gstack-telemetry-log" --event-type onboarding --skill _setup_playwright --outcome "$_PW_FAIL_REASON" --no-sweep >/dev/null 2>&1 || true
fi
fi
if [ ${#_FOREIGN_SKIPPED_ENTRIES[@]} -gt 0 ]; then
log ""
log "Not registered (a skill you own already uses the name; left untouched): ${_FOREIGN_SKIPPED_ENTRIES[*]}"
log " Rename or move yours, or switch modes (./setup --prefix / --no-prefix) so the names no longer collide."
fi
if [ ${#_BACKED_UP_SKILL_MDS[@]} -gt 0 ]; then
log ""
log "Moved ${#_BACKED_UP_SKILL_MDS[@]} customized SKILL.md file(s) to $_SKILL_BACKUP_ROOT before installing gstack's: ${_BACKED_UP_SKILL_MDS[*]}"
log " Those were gstack-generated files you had edited; restore anything you meant to keep under a different skill name."
fi