docs(agents): use precise terminology for action.yml heading

[fullsend-ai-coder]

Summary

• Rename the bold heading on AGENTS.md line 94 from "Action workflows (action.yml)" to "Composite action (action.yml)"
• In GitHub Actions, "workflows" refers to .github/workflows/*.yml files, while action.yml defines a composite action
• The old heading introduced the same class of terminological ambiguity that the repo's "Terminology: tier conventions" section exists to prevent

Closes #3282

Test plan

• Verified AGENTS.md passes pre-commit lint checks
• Verified no cross-references to the old heading text exist elsewhere in the repo
• Confirmed CLAUDE.md and other docs do not reference the old heading

[github-actions]

E2E tests did not run

E2E tests run automatically for org/repo members and collaborators on pull requests.

For other contributors, a maintainer must add the ok-to-test label after the latest push.

See E2E testing guide for details.

[github-actions]

E2E tests did not run

E2E tests run automatically for org/repo members and collaborators on pull requests.

For other contributors, a maintainer must add the ok-to-test label after the latest push.

See E2E testing guide for details.

[github-actions]

Site preview

Preview: https://df552c96-site.fullsend-ai.workers.dev

Commit: bc3652e776ed2e3de1adcfe550d8d085fcd6d263

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

[fullsend-ai-retro]

🤖 Finished Retro · ✅ Success · Started 10:47 PM UTC · Completed 11:00 PM UTC
Commit: bc3652e · View workflow run →

[fullsend-ai-retro]

PR #3286 was a clean, well-executed 1-line terminology fix that renamed the "Action workflows (action.yml)" heading to "Composite action (action.yml)" in AGENTS.md. The code agent produced a single commit with zero rework. The retro→code pipeline worked correctly end-to-end: the retro agent on PR #2219 caught an unaddressed info-level review finding about ambiguous GitHub Actions terminology → filed issue #3282 → code agent created PR #3286 → ralphbean approved → merged same day.

Two inefficiencies stand out:

1. Multi-hop feedback loop for a 1-word change: The review agent on PR docs(#2215): extend forge abstraction scope to action.yml #2219 correctly flagged the terminology issue as [info] severity. Human reviewers approved without addressing it. The retro then had to catch it post-merge, spinning up the full pipeline (retro analysis + issue creation + code agent PR + human review + CI). Total overhead: 3+ agent runs, ~12 hours latency, for what could have been addressed inline during PR docs(#2215): extend forge abstraction scope to action.yml #2219's review cycle.

2. No review agent ran on PR docs(agents): use precise terminology for action.yml heading #3286: The dispatch job completed in 7 seconds with the 'Determine stage' step producing no stage — all subsequent steps were skipped. This confirms the bot authorization gap tracked in #2674: is_event_actor_authorized("fullsend-ai-coder[bot]") fails because GitHub App bots aren't collaborators. PR docs(agents): use precise terminology for action.yml heading #3286 modified AGENTS.md (a protected governance file) with zero automated review. This is the same pattern documented in dispatch: ADR 0054 authorization gate blocks review dispatch for coder-bot PRs #2674 and Prioritize review dispatch for bot-authored PRs that modify agent definitions #2988.

Existing issue coverage for patterns observed:

• dispatch: ADR 0054 authorization gate blocks review dispatch for coder-bot PRs #2674 — dispatch authorization gate blocks review for bot PRs (directly explains missing review on PR docs(agents): use precise terminology for action.yml heading #3286)
• Prioritize review dispatch for bot-authored PRs that modify agent definitions #2988 — urgency for bot PRs modifying agent definitions
• Review agent medium findings should trigger fix agent or block merge #870 — medium findings should trigger fix agent or block merge (adjacent; if implemented with the severity escalation below, would prevent the roundtrip)
• Retro agent should apply a higher bar before filing issues from low-severity doc-only findings #2949 — retro should apply higher bar for low-severity doc-only findings (partially relevant, though this finding was genuinely valid)

One novel proposal below addresses the severity rating gap that caused the multi-hop feedback loop.

Proposals filed

• Review agent should escalate terminology-precision findings on governance docs with explicit terminology conventions (in fullsend-ai/fullsend)