Skip to content

ci: bump third-party action versions across workflows + SHA-pin pr-checks.yml - #69

Merged
heskew merged 1 commit into
mainfrom
workflow/bump-third-party-actions-sweep
May 4, 2026
Merged

ci: bump third-party action versions across workflows + SHA-pin pr-checks.yml#69
heskew merged 1 commit into
mainfrom
workflow/bump-third-party-actions-sweep

Conversation

@heskew

@heskew heskew commented May 4, 2026

Copy link
Copy Markdown
Member

Summary

Two things in one PR — they're both "all third-party actions on this repo to current SHA-pinned versions past the 3-day buffer":

Concern Files Detail
Bump to current versions claude-*.yml, auth-gate-invariants.yml, release.yml Same three bumps as harper#464 — `actions/checkout`, `actions/setup-node`, `anthropics/claude-code-action`. Plus mirror of harper#461's checkout v4.3.1 → v6.0.2 (which never came across to oauth).
SHA-pin tag-pinned refs pr-checks.yml Three references were carrying supply-chain risk via mutable tags: `actions/checkout@v4`, `actions/setup-node@v4`, `oven-sh/setup-bun@v2`. All now pinned to the same SHAs the rest of the repo uses.

Bumps

Action Was Now Released Age
`actions/checkout` v4.3.1 v6.0.2 2026-01-09 4 months
`actions/setup-node` v6.2.0 v6.4.0 2026-04-20 14 days
`anthropics/claude-code-action` v1.0.99 v1.0.110 2026-04-29 5 days

SHA pins (pr-checks.yml)

Was (tag-pinned) Now (SHA-pinned)
`actions/checkout@v4` `@de0fac2e... # v6.0.2`
`actions/setup-node@v4` `@48b55a01... # v6.4.0`
`oven-sh/setup-bun@v2` `@0c5077e5... # v2.2.0`

`actions/setup-node` went `v4 → v6` in pr-checks.yml. v5 was the Node 20 → Node 24 jump; v6 minor releases since. Backwards-compatible for typical setup-node usage.

Held back

  • `anthropics/claude-code-action` v1.0.111 (2026-05-01) — at the 3-day buffer edge today; deliberately one patch behind. Will pick up on the next sweep.

Verified locally

  • All affected workflows parse as valid YAML.
  • `bash .github/scripts/validate-auth-gate-invariants.sh` passes against all three `claude-*.yml` workflows.

Test plan

  • Push to a PR after this lands → confirm `Auth gate invariants / validate` still passes.
  • Confirm `pr-checks` workflow still runs cleanly with v6 setup-node.
  • Confirm a Claude review run completes end-to-end on v1.0.110.

🤖 Generated with Claude Code

…ecks.yml

Sweep of currently-pinned third-party action SHAs to latest stable
releases past the 3-day buffer, plus SHA-pin three tag-pinned
references in pr-checks.yml that were carrying supply-chain risk.

Bumps applied:

- **actions/checkout**: v4.3.1 → v6.0.2 (mirror of HarperFast/harper#461).
  v5.0.0 was the Node 20 → Node 24 jump; v6.0.2 (2026-01-09) is
  the latest stable, ~4 months baked. 11 instances across:
  - auth-gate-invariants.yml (1)
  - claude-issue-to-pr.yml (3)
  - claude-mention.yml (3)
  - claude-review.yml (4)

- **actions/setup-node**: v6.2.0 → v6.4.0 (2026-04-20, 14 days
  old). Minor version. 3 instances in claude-issue-to-pr.yml,
  claude-mention.yml, release.yml.

- **anthropics/claude-code-action**: v1.0.99 → v1.0.110
  (2026-04-29, 5 days old). Patch range; v1.0.111 (2026-05-01) is
  at the buffer edge so conservatively held back. claude-issue-
  to-pr.yml, claude-mention.yml, claude-review.yml.

SHA pins added to pr-checks.yml (was tag-pinned, supply-chain risk):

- `actions/checkout@v4`           → @de0fac2e... # v6.0.2
- `actions/setup-node@v4`         → @48b55a01... # v6.4.0
- `oven-sh/setup-bun@v2`          → @0c5077e5... # v2.2.0

Note: actions/setup-node went v4 → v6 in pr-checks.yml. v5 was
the Node 20 → Node 24 jump; v6 minor releases since. Backwards-
compatible for typical setup-node usage.

Skipped (already current past buffer):
- actions/create-github-app-token v3.1.1 (2026-04-11)
- oven-sh/setup-bun v2.2.0 (2026-03-14, 51 days old)

Verified locally:
- All affected workflows parse as valid YAML.
- `validate-auth-gate-invariants.sh` passes against all three
  claude-*.yml workflows.

Continues parallel to harper#464.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
@heskew
heskew requested a review from a team as a code owner May 4, 2026 16:24
@heskew
heskew merged commit c975a7b into main May 4, 2026
7 checks passed
@heskew
heskew deleted the workflow/bump-third-party-actions-sweep branch May 4, 2026 16:25
@claude

claude Bot commented May 4, 2026

Copy link
Copy Markdown
Contributor

Reviewed; no blockers found.

heskew added a commit that referenced this pull request Jun 26, 2026
…on + log-count fix) (#135)

Moves the claude-review (was 9eb635a, week-of-06-08) and gemini-review
(was 48e539a, post-#66) callers up to ai-review-prompts main 1bbc562:

- #67 — week-of-06-15 calibration: "error/abort paths that crash or leak"
  Robustness rule; "explicitly-throwaway / empty-diff PRs" ignore; the
  mechanical-diff ignore narrowed to inert CI-config with a workflow-
  security carve-out.
- #69 — log-review finding-count fix: a finding-bearing review is no
  longer logged to ai-review-log as "no blockers" when the count is
  spelled out or inline-only (ai-review-log #683).

uses: and ai-review-prompts-ref: move in lockstep. Part of the harper /
harper-pro / oauth rollout of this batch.

Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
heskew added a commit that referenced this pull request Jun 29, 2026
…t-ref tracking #71) (#136)

* chore(ci): bump ai-review-prompts to 67d7611 (prompt-ref tracking)

Forward bump 1bbc562 -> 67d7611 (ai-review-prompts #71): the log step now
records which prompt version produced each ai-review-log entry — a
`**Prompt ref:**` body field plus a `prompt:<shortsha>` label — so
calibration can attribute verdicts to a specific prompt version instead
of a date bucket. No prompt-content change vs 1bbc562 (which already
carried the #67 calibration + #69 log-count fix); this turns on
per-version tracking.

`uses:` and `ai-review-prompts-ref:` move in lockstep.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

* chore(ci): re-point bump to 9cf49d2 (add #70 week-of-06-22 calibration)

#70 (week-of-06-22 calibration) merged after this bump was opened, so
re-point 67d7611 -> 9cf49d2 to land the calibration AND the #71 prompt-ref
tracking together in a single bump rather than forcing a second one.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant