Closes#410, #2037, #2045, #1640, #1280, #1329, #1341, #255,
anthropics/claude-code#46720 (partial closes on overlap with other rules).
The plugin's substring-only XSS / browser-DOM rules
(new_function_injection, react_dangerously_set_html, document_write_xss,
innerHTML_xss, outerHTML_xss, insertAdjacentHTML_xss) fired on any file
containing the trigger substring — including:
* Markdown documentation explaining XSS sinks
* Blog posts / READMEs that name browser APIs
* Python tutorials referencing dangerouslySetInnerHTML
* Plugin skill files with example HTML strings
* .yaml / .json configs that happen to contain the literal string
* .gitignore / Dockerfile / Makefile
These constructs have no meaning outside JS/TS source. Add a
path_filter: lambda p: p.endswith(_JS_EXTS) to each so they fire only
on .js, .jsx, .ts, .tsx, .mjs, .cjs, .mts, .cts, .vue, .svelte.
Cross-checked against the existing _JS_EXTS-gated rules
(regex_exec_substring, child_process_exec, exec_substring) — same
pattern, same constant, same intent. Uses the module-level _JS_EXTS
tuple so future extension changes propagate to all 6 rules atomically.
Verified locally on macOS Python 3.13:
- py_compile clean.
- 45-test existing smoke + extensibility suite still passes.
- 151 new parametrized tests in test_xss_gate.py (added to internal
test suite this PR doesn't ship): each gated rule x every
JS-family extension accepts, x every non-JS path (.md / .py /
.yaml / .json / .txt / .html / Dockerfile / Makefile / .gitignore
/ .sh / .go / .rs / .rb) rejects. 196 tests pass total.
Doesn't address everything in the false-positive cluster — issues that
require Python-rule gating (#1114 .env.schema exec), tighter substring
scoping (#660 pickle in usernames), or hook-protocol changes (#1358
exit-2 vs warning, #1375 plain-text-vs-JSON output) need separate PRs.
This PR covers the JS-substring subset cleanly.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Fixes anthropics/claude-plugins-official#2071 — on macOS where the
default `python3` is Apple's Command Line Tools Python 3.9.6, the
plugin's agentic commit reviewer silently does not run, even when the
user has a newer Python installed.
Three compounding factors in the bug:
1. `sg-python.sh` only checks the major version (`3`), so it always
picks 3.9 even when 3.10+ is on PATH.
2. `claude_agent_sdk` requires Python >=3.10 — pip install on 3.9
returns "No matching distribution" -> bootstrap returns BUILD_FAILED.
3. Even with a hand-built 3.12 venv, `llm.py` imports the SDK
in-process into the hook's interpreter (still 3.9), which raises
SyntaxError. The existing venv-probe in `ensure_agent_sdk.py` uses
the venv's own Python (3.12) so it reports NOOP_VENV (healthy) while
the consumer fails — misleading telemetry on top of silent feature
degradation.
Per BQ telemetry, 14,073 external macOS users hit
sdk_bootstrap=BUILD_FAILED in the past 4 days (the default-macOS
cohort), out of ~86K total external installed users. Combined with
~20K other users in similar broken-bootstrap states (Windows pre-#2055,
Linux <3.10), about half the installed base has a silently-broken
agentic reviewer.
This PR implements the reporter's items #1, #3, and #4. Item #2
(running the SDK out-of-process) is deferred as a bigger refactor.
Item #1 — hooks/sg-python.sh — prefer >=3.10 binaries via 3-pass probe:
Pass 1: python3.13 / 3.12 / 3.11 / 3.10 (>=3.10 by name, highest wins)
Pass 2: bare python3 / python / py -3 (accept only if reported >=3.10)
Pass 3: bare python3 / python / py -3 (any Python 3, FALLBACK so
pattern checks still work on macOS-default 3.9 — no regression
vs today; SDK-dependent paths detect the version mismatch
inside Python and degrade cleanly via item #4)
Item #4 — ensure_agent_sdk.py — health-check honesty:
Added HOOK_PY_INCOMPATIBLE=6 outcome with short-circuit at top of main():
if sys.version_info < (3, 10):
return HOOK_PY_INCOMPATIBLE, "hook_py", f"py_{...}"
Telemetry consequences after rollout: sdk_bootstrap=6 is a new clean
bucket; some users currently miscounted in sdk_bootstrap=3 BUILD_FAILED
(wasted pip cycles) and sdk_bootstrap=1 NOOP_VENV (falsely-healthy)
move to sdk_bootstrap=6. The remaining NOOP_VENV count becomes
trustworthy.
Item #3 — ensure_agent_sdk.py — one-time user-visible notice:
When outcome == HOOK_PY_INCOMPATIBLE and a marker file at
`~/.claude/security/.agentic_unavailable_notice_v<pv>` doesn't exist,
the SessionStart response includes hookSpecificOutput.additionalContext
+ systemMessage explaining the situation. Marker file is plugin-
version-keyed so a future fix (e.g. shipping out-of-process SDK) can
bump pv and re-notify users.
BUILD_FAILED is intentionally excluded from the notice — it covers
transient causes where a permanent banner would mislead.
Verified locally on macOS Python 3.13:
- py_compile clean on both files.
- Existing 45-test smoke + extensibility suite: 45/45 PASS in 2.50s.
- Unit test of simulated 3.9 path: HOOK_PY_INCOMPATIBLE returned with
correct phase/kind; notice shown on first call, suppressed on
second, reshown on bumped pv; BUILD_FAILED correctly does NOT
trigger notice.
NOT verified: actual Python 3.9 behavior end-to-end (would need a 3.9
install). Worth a follow-up smoke test in a 3.9 venv before next
release. The unit test simulating 3.9 covers the logic but not the
runtime invocation through the shim.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Adds an `emit-verdict` job to scan-plugins.yml that posts a sticky
comment per scanned entry to the corresponding bump PR, with marker
`<!-- bump-pr-verdict:<name> -->`. The body is a schema_v1 JSON block,
the same shape `anthropics/claude-plugins-community-internal`'s
`scan-external-plugins.yml` already emits, so any consumer that already
reads verdicts from that schema works uniformly across both repos.
What this enables
-----------------
Lets downstream consumers (label automation, dashboards, anything that
wants per-entry verdict signal) read verdicts directly from the PR
rather than scraping job logs or downloading artifacts. The current
options are log-scraping (truncated after log retention) or fetching
the `scan-verdicts` artifact (retention-limited and only after upload
succeeds).
What does NOT change
--------------------
- The `scan` required check is unaffected (emit-verdict is
`continue-on-error: true` at the job level — failures here MUST NOT
block the required gate).
- Verdict cache, scan flow, and revert-failed-bumps.yml are unchanged.
- No new permission scopes (uses `pull-requests: write` at the job
level, identical to other PR-commenting jobs in this repo).
Schema notes
------------
- `scan.*` axes (clone, schema, binaries, etc.) emit as "skipped" —
this workflow runs the policy review only, not per-entry static
checks. Shape kept compatible with -internal's schema_v1 so the
same consumers work uniformly on both repos.
- `policy.has_broad_scope_hooks`, `has_undisclosed_telemetry`,
`description_matches_behavior` emit as null — those granular axes
aren't surfaced by this workflow's per-entry artifact yet. Consumers
that map `null → "?"` for display already handle this gracefully.
- `policy.status` is execution state (not outcome). Map source →
status: scan-action-run → "ran"; cache-served → "cached". Outcome
lives in `policy.passes`. policy.status vocabulary matches the
`ran|cached|missing|gated_out|infra_error` convention from
-internal's emit-verdict.
PR resolution
-------------
`pull_request` events carry the PR number directly. The bump workflow
creates bump PRs via GITHUB_TOKEN (which doesn't fire `pull_request`
triggers — recursion guard) and dispatches this scan via
`workflow_dispatch` on the bump branch; in that case the job looks up
the open PR by head ref via REST. No PR found (scan_all dispatch on
main, etc.) → no-op with notice.
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Each internal plugin already carries an Apache 2.0 LICENSE file at its root.
This adds the same license at the repository root for clarity.
Co-authored-by: Claude <noreply@anthropic.com>
Picked up from sethconvex's PR #1966 (auto-closed by membership gate),
split off from #1980 (Convex plugin entry refresh) so the editorial
addition to claude-automation-recommender gets its own review.
Changes:
- SKILL.md: add `convex` to the package.json dep-detection grep, update
the Database row in the indicator table to name Convex, and add a
Convex MCP row to the MCP recommendation table.
- references/mcp-servers.md: new "Convex MCP" section in the Databases
group (Supabase / Convex / PostgreSQL / Neon / Turso), and a row in
the Detection Patterns quick reference.
Convex publishes its MCP server via the `convex` npm package
(`npx convex mcp start`), exposing tables, function-spec, data,
run-once-query, logs, env list/set/get. Same row pattern as the
existing database/backend MCP entries.
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* Refresh Convex plugin: rename to `convex`, bump SHA to v1.0.1, richer metadata
Picked up from sethconvex's PR #1966 (auto-closed by membership gate).
Original entry added by Tobin in PR #1918 (2026-05-18).
Changes to the Convex marketplace.json entry:
- **Rename slug** `convex-backend` → `convex` to match the single-brand-word
convention used by every peer in the database/backend neighborhood
(`supabase`, `firebase`, `mongodb`, `prisma`, `clickhouse`, `cockroachdb`,
`cloud-sql-postgresql`, `alloydb`). New `displayName: "Convex"` keeps the
directory UI label unchanged.
- **Bump SHA pin to `59663a5`** (plugin v1.0.1) — current HEAD of
`get-convex/convex-backend-skill` `main`. New SHA adds:
- `agents/convex-expert.md` — subagent encoding non-negotiable Convex code
rules (object-form syntax, validator requirements, index naming,
internal-vs-public, schema evolution, resource limits). Loaded only
when delegated to.
- `monitors/monitors.json` — runtime-error monitor streaming
`npx convex logs`, surfacing matched errors as notifications. Self-guards
on unlinked projects. `when: on-skill-invoke:design` so it only starts
after the skill is invoked.
- `.mcp.json` — auto-wires the Convex MCP server
(`npx -y convex@latest mcp start`, local stdio).
- Public-facing README (install / how-to-use / what's bundled / capabilities).
- `paths` gate on the skill — `[convex/**, convex.json, package.json]` for
auto-invocation precision.
- `description` / `when_to_use` split on the skill frontmatter.
- **Refresh marketplace entry metadata** — `displayName`, `keywords` (15
discovery tags), `author.url`, expanded `description`, category changed from
`development` to `database` (matches every peer), `homepage` repointed at the
plugin repo (matches the `supabase` pattern).
Verified locally:
- Author affiliation confirmed: `seth@convex.dev` commit email, write access
to the canonical `get-convex/` org.
- `claude plugin validate`: PASS.
- Static audit: PASS @ 92 (manifest 96, security 93, quality 80, docs 100).
- MCP server is local stdio (`has_remote_mcp=false`) — passes the -official
add-official Phase 2e gate.
Recommender skill changes from the original PR are split into a follow-up.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* Re-pin Convex to 5e59870 (post upstream fix merge)
Upstream PR get-convex/convex-backend-skill#1 merged 2026-05-23. The
agents-field array-shape fix now applies; claude plugin validate passes
on both the full plugin (with marketplace.json) and the isolated
plugin.json — including the external-validator gate this PR previously
failed on.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The first round of this PR removed SKIP_WIN32, fixed venv_py to use
Scripts/python.exe, and added Lib/site-packages to the consumer glob —
all necessary. Windows verification (Win11 ARM64, Py 3.13, Git Bash)
showed two more blockers, both addressed here.
1. Pip dependency resolver picks unbuildable cryptography on ARM64.
Without --prefer-binary, pip picks a cryptography version with no
published ARM64 wheel and tries to build it from source. That needs
Rust/Cargo, almost never present on user machines → BUILD_FAILED
with err_kind=other:cryptography. A binary wheel exists for an
adjacent version (cryptography-46.0.3-cp311-abi3-win_arm64.whl);
--prefer-binary tells pip to pick it. Cross-platform safe (no-op
where the latest version already has a wheel).
2. pywin32 .pth files aren't processed by sys.path.insert().
With the venv built, ensure_agent_sdk.py's post-build probe passes
(it runs from venv_py, where Python's site.py at startup processes
pywin32.pth and registers win32/, win32/lib/ plus runs
pywin32_bootstrap.py to set the DLL search dir). But llm.py runs in
the hook's SYSTEM Python and adds the venv via sys.path.insert(),
which doesn't trigger site.py at all. Without the bootstrap, the
SDK's mcp.client.stdio → mcp.os.win32.utilities chain raises
ModuleNotFoundError: pywintypes and the agentic reviewer falls back
to single-shot silently — exactly the symptom this PR is trying to
fix. The probe says NOOP_VENV; the actual consumer fails. Probe and
consumer use different Pythons.
Replicate what site.py would do: after inserting site-packages,
also insert win32/ and win32/lib/, then exec pywin32_bootstrap.py.
Pulled into a shared helper _inject_agent_sdk_venv_into_syspath()
so both consumer sites (3P SDK fallback, agentic_review fallback)
call the same code — Windows handling stays in one place.
Verified on macOS (POSIX path unchanged):
- Helper end-to-end test: POSIX-layout venv detected + fake package
imports successfully via the injected path
- Windows-layout venv also detected; win32 branch correctly skipped
via sys.platform check
- Both files pass py_compile
Credit: @mhegazy verified the previous commit on Win11 ARM64 / Py 3.13
/ Git Bash, surfaced both issues end-to-end, and provided the exact
fix patterns. This commit applies them with the pywin32 part factored
into a shared helper (vs. inlining at both consumer sites).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The agentic reviewer is silently no-op on Windows today. SessionStart
bootstrap (ensure_agent_sdk.py) short-circuits with SKIP_WIN32 because
the consumer glob in llm.py only matches POSIX venv layout
(lib/pythonX.Y/site-packages). On Windows, venvs use Lib/site-packages
(capital L, no pythonX.Y subdir), so even if a venv got built the
glob wouldn't find its contents.
Result: Windows users on default installs (no system-wide
claude_agent_sdk) get layer 1 (pattern warnings) and layer 2 (single-
shot LLM diff review) but not layer 3 — the cross-file agentic review
that catches IDOR, auth-bypass, cross-file SSRF, and other things that
need to read related files. Plugin description claims layer 3 but it
silently doesn't run.
Three changes:
1. llm.py — extend the consumer glob (2 sites: 3P SDK fallback at
~L297, agentic_review fallback at ~L1090) to also match the Windows
Lib/site-packages layout, so a venv built on Windows is actually
discoverable.
2. ensure_agent_sdk.py — remove the sys.platform == 'win32' early-exit
so the SessionStart bootstrap builds the venv on Windows too.
Outcome code 4 (formerly SKIP_WIN32) is retired but not reused so
pre-fix telemetry rows still decode correctly.
3. ensure_agent_sdk.py — venv_py path now branches on sys.platform:
Windows venvs put the interpreter at Scripts\python.exe; POSIX
uses bin/python. Previously assumed POSIX, so even with the glob
fix, the post-build SDK-importability probe would fail on Windows.
Verified locally on macOS:
- glob test: both layouts now match (POSIX venv detected, simulated
Windows venv also detected via the new Lib/site-packages branch)
- both files pass py_compile
- POSIX path unchanged (sys.platform != 'win32' so old branch runs)
Not verified on Windows in this commit — needs an actual Windows
runner to confirm the venv build + SDK import + subprocess plumbing
all work end-to-end. The SDK spawns a child claude.exe; Windows
process plumbing has its own quirks (shell semantics, path escaping)
that may surface separately. Worth a controlled rollout (one-week
soak under env-var opt-in before flipping default).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Fixes#2043. On Git Bash for Windows, Claude Code hands script paths to
the shim in POSIX form (`/c/Users/...`). We exec a Windows `python.exe`
(the `python3` Microsoft Store stub fails the probe), and Windows Python
interprets the leading `/` as the root of the current drive — `/c/...`
becomes `C:\c\Users\...` or `D:\c\Users\...` depending on which
drive the shell happens to be on, fails with ENOENT, and every
Edit/Write/MultiEdit blocks until the session restarts.
Convert absolute path args via `cygpath -w` (a Git Bash builtin) before
exec. Guarded by `command -v cygpath` so macOS/Linux fall straight
through unchanged; `cygpath -w` is idempotent on already-Windows paths
so the rare mixed-form case is safe. Only `/*` paths are converted —
Windows-form paths reaching the shim are already openable by python.exe.
Verified locally:
- cygpath absent on macOS → guard skips → POSIX behavior unchanged
- end-to-end shim invocation with a POSIX path on macOS exits 0
- stubbed cygpath -w on /c/Users/test/hook.py produces C:\Users\test\hook.py
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* ci: migrate scan-plugins.yml to Workload Identity Federation auth
Replaces the static ANTHROPIC_API_KEY repo secret with Workload
Identity Federation: the scan-plugins shared action mints a GitHub
OIDC token (id-token: write) and the claude CLI exchanges it for a
short-lived bearer. The federation rule is bound to this repository
(repository_id-pinned).
Depends on anthropics/claude-plugins-community#34 (adds the WIF
inputs to the shared action). Pinned to that PR's head SHA; will
re-pin to a main-branch SHA once #34 merges.
Drops the 'Require ANTHROPIC_API_KEY' fail-closed guard — the WIF
inputs are literal in this file, so the action's skip-if-no-auth
path can't trigger. Updates the prompt-injection security comment
to reflect the short-lived bearer model.
* scan-plugins: re-pin to cpc#34 merge commit on main
claude-plugins-community#34 merged at e85f0d65b4fc87f07862e1dcdc467950514414ec — re-pinning from
the PR head SHA to the squash-merge commit on main so the pin survives
any future branch GC.
Updates ui5 and ui5-typescript-conversion to the renamed upstream
repo UI5/plugins-coding-agents (formerly UI5/plugins-claude) and
bumps both SHA pins to current upstream main.
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
curl writes "000" to -w '%{http_code}' on a connection failure AND exits
nonzero. The previous fallback put the echo inside the command
substitution — both wrote, the captured value was "000000", and the
case statement's 000) arm didn't match, so dead hosts fell through to
PASS. Move the fallback assignment outside the substitution so the
captured value is exactly "000" and connection failures fail.
Also skip entries with an empty url field — those are placeholders
awaiting user config, not dead endpoints, and would false-fail.
Aligns the compose service name, local config filename, and all
log/restart commands with the image and binary name. Adds an explicit
-config arg since the image CMD still defaults to the legacy
/etc/mcp-gateway path.
🏠 Remote-Dev: homespace
- Replace doc references with platform.claude.com URLs (overview,
quickstart, security, deploy-compose, deploy-helm, console,
troubleshooting, reference, WIF)
- Swap the POC mcp-proxy image for the public registry digest used in
the published quickstart
🏠 Remote-Dev: homespace
Adds the /create-docker-mcp-tunnel command, which drives the MCP tunnels
Docker Compose quickstart end to end: preflight checks, certificate
generation, proxy config, cloudflared, an optional sample FastMCP server,
and verification from Managed Agents and the Messages API.
Migrated from anthropic-experimental/mcp-tunnel-skills.
🏠 Remote-Dev: homespace
Bumps the mercadopago plugin pin from 1de8d97e to 63ff263c (latest main).
v2 replaces the mcp-launcher.sh keychain-read / npx -y mcp-remote
wrapper with a plain type:"http" MCP entry pointing at
https://mcp.mercadopago.com/mcp, and consolidates 13 skills into 4
orchestration skills. The pinned SHA also includes the May 19 fix
that gates the PreToolUse hook on project relevance so it no longer
runs on unrelated projects.
Description updated to match the partner's v2 self-description.
https://claude.ai/code/session_01KRC2Uv6UaFFdrt7sjn45yT
Co-authored-by: Claude <noreply@anthropic.com>
Paths containing spaces (common on Windows, e.g. C:\Users\Some User\...)
cause shell word-splitting when CLAUDE_PLUGIN_ROOT is unquoted, resulting
in hooks erroring with "No such file or directory" on every tool call.
Wraps the path in double quotes for all five affected hook commands.
Fixes the pattern reported in issue #57946. Closes the fix surfaced in PR #1921.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* Bump 26 plugin SHA pins to upstream HEAD
* Revert mercadopago SHA bump
The new upstream SHA adds a PreToolUse hook that fires on every
Bash/Edit/Write/Read in all sessions and globally blocks reading .env
files, regardless of project relevance. The policy scan flags this as
out of scope for what the plugin description advertises. Leave at the
prior pin until the upstream gates the hook on project relevance.
* Fix broken plugin source configs and bump their SHAs
Several external plugins had source configs that no longer matched the
upstream layout, so the automated SHA bump skipped them indefinitely.
Add the missing path field where the manifest moved into a subdirectory,
correct stale ref/commit metadata, and update the skills list for the
one strict:false skills-only entry.
- rc, revenuecat: upstream moved the plugin from repo root into
revenuecat/. Add path and bump SHA.
- zilliz: plugin moved from repo root into plugins/zilliz/. Add path
and bump SHA.
- sumup: plugin lives at providers/claude/plugin/ (declared by the
upstream marketplace.json) but our entry never had a path. Add it
and bump SHA.
- mintlify: pure SHA bump. Repo layout unchanged between SHAs; the
upstream remains a marketplace-style repo with no plugin.json, same
as the currently pinned SHA.
- netsuite-suitecloud (strict:false skills entry): bump SHA and add
the four new skill directories upstream added since the last pin.
- 42crunch-api-security-testing: ref said v1.0.1 but the pinned SHA
is actually v1.5.5. Correct the label; the SHA is already current.
- jfrog: commit and sha fields had drifted apart. Set both to
upstream HEAD.
Each new SHA verified to be on the upstream default branch and the
referenced manifest validated with claude plugin validate.
* Revert mintlify and netsuite-suitecloud changes
The validate-plugins check requires a plugin manifest at the pinned SHA
even for strict:false entries. Neither repo has one at any SHA, so a
SHA bump fails CI. Leave them at the existing pin until either the
upstream adds a manifest or the validator learns to honor strict:false.
* Cache scan verdicts and drop policy-failing entries from bump PRs
Three changes that together let the nightly bump clear any backlog in a
single run without blocking on a single bad upstream or re-burning Claude
time on already-scanned SHAs:
- bump-plugin-shas.yml: raise max-bumps default 20 -> 130 (above the
external entry count, so a single run can clear a full backlog) and add
an explicit 60-min job timeout. The cap was the only thing bounding the
blast radius of a single policy failure; the changes below take over
that role so the cap can be lifted.
- scan-plugins.yml: add a verdict cache keyed on (plugin, sha, policy
hash). The bump action force-resets bump/plugin-shas every night, which
makes the same SHAs reappear in the diff on consecutive nights — without
the cache the scan would re-burn ~90s of Claude time per entry per
night. Cached verdicts (pass and fail) are served from disk; only
uncached SHAs are scanned. The job still fails on cached failures so
the required check stays honest.
- revert-failed-bumps.yml (new): after a Scan Plugins workflow_run on
bump/plugin-shas concludes with a failure, drop just the failing
entries' source.sha back to main's pin via a follow-up signed commit
and re-dispatch the scan. The re-dispatch finds only cached-pass
entries and goes green in seconds. Bounded at 3 passes/night, restricted
to SHA-only diffs, and aborts if the bump branch was tampered with.
* Harden bump cache and revert workflows after review
- revert-failed-bumps: replace the time-based revert budget (anchored on
the PR head, which a revert commit immediately replaces — never
accumulating past 1) with a commit count: every nightly bump force-
resets to one commit and every revert pass adds exactly one, so
commits > MAX+1 is the budget without date math, pagination, or
exposure to comment spoofing.
- revert-failed-bumps: filter the bump PR by head owner so a fork PR
with a branch named bump/plugin-shas can't be selected.
- revert-failed-bumps: continue-on-error on the artifact download so a
scan that died before uploading (infra error) doesn't fail the revert
job — the missing-file guard downstream handles it.
- scan-plugins: add a per-ref concurrency group so concurrent scans
don't lose one another's cache writes; key the cache on run_attempt
so a re-run can save its own verdicts.
- scan-plugins: store the full source object in the cache and require
source equality on lookup, so a repo/path change at the same SHA
misses the cache instead of getting a stale verdict.
- scan-plugins / revert-failed-bumps: strip markdown control chars,
wrap model-generated text in code spans (neutralizes auto-linked
URLs), and redact key-shaped tokens before they reach the step
summary, artifact, cache, or PR comment.
Upstream plugins move daily; a weekly sweep with a 20-bump cap can fall
behind. Each run force-resets the bump branch, so stale unmerged PRs are
replaced rather than piling up.
Walks marketplace.json for vendored plugins, extracts http/sse MCP
server URLs from .mcp.json / mcp.json / plugin.json, and probes each
with HEAD then a JSON-RPC POST fallback. Fails on 404/410 and
connection errors; passes on auth/method errors (expected without
credentials). Runs on PR, daily schedule, and manual dispatch.
External (SHA-pinned) plugins are out of scope — their .mcp.json
isn't checked out here.
