Skip to content

docs(#7283): drop ADR skill line-count hard stops - #7285

Merged
waynesun09 merged 1 commit into
mainfrom
agent/7283-relax-adr-line-limit
Sep 14, 2026
Merged

waynesun09 merged 1 commit into
mainfrom
agent/7283-relax-adr-line-limit

Conversation

@fullsend-ai-coder

Copy link
Copy Markdown
Contributor

Summary

Remove the ADR writing skill's numeric length hard stops so authors are not pressured to compress a decision that needs more room. Conciseness still comes from one decision per ADR, linking to problem docs instead of restating them, and short Consequences bullets.

Related Issue

#7283 — Relax the ADR writing skill’s 100-line hard limit

Changes

  • Drop the "Total ADR length: Aim for under 80 lines / over 100" bullet from the Be concise section
  • Drop the "ADR is over 100 lines of content" red flag
  • Leave remaining conciseness guidance and the per-section target table unchanged

Testing

  • Secret scan passed on the changed skill file
  • Applicable markdown hooks ran directly (trailing whitespace, EOF, merge conflict markers, line endings, gitleaks, broken-symlink lint, lychee)
  • make lint could not run as a whole: pre-commit failed to fetch remote hook repos (HTTP 403)
  • No production code or tests to update (docs/skill text only)

Checklist

  • PR title follows Conventional Commits (docs(#7283): ...)
  • Autonomous agent commit (DCO waived for bot identity)
  • I wrote this contribution myself and can explain all changes in it

Closes #7283

Post-script verification

  • Branch is not main/master (agent/7283-relax-adr-line-limit)
  • Secret scan passed (gitleaks — b95c0b486f88df696ef3c0d6d53577cd55af9ea0..HEAD)
  • PR body secret scan passed (gitleaks — no-git)

The writing-adrs skill treated "under 80 lines" and "over 100
lines" as a hard length limit, which pressured authors to
compress ADRs even when the decision needed more room.

Remove those numeric thresholds. Existing conciseness guidance
(one decision per ADR, link instead of restating problem docs,
short Consequences bullets) is enough.

Closes #7283
@fullsend-ai-coder fullsend-ai-coder Bot added the ready-for-review Triggers review agent dispatch label Sep 14, 2026
@fullsend-ai-coder
fullsend-ai-coder Bot requested a review from a team September 14, 2026 14:10
@fullsend-ai-review

Copy link
Copy Markdown

🤖 Review · ⚠️ Cancelled · Ended 2:11 PM UTC

Commit: 902a609 · View workflow run →

@fullsend-ai-review

fullsend-ai-review Bot commented Sep 14, 2026 •

Copy link
Copy Markdown

🤖 Finished Review · ✅ Success · Started 2:13 PM UTC · Completed 2:24 PM UTC

Commit: 902a609 · View workflow run →

Runtime: pi · Model: sonnet → claude-sonnet-5 · Effort: high · Cost: $1.46

@codecov

codecov Bot commented Sep 14, 2026

Copy link
Copy Markdown

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

@fullsend-ai-review

Copy link
Copy Markdown

Review

Findings

Medium

  • [protected-path] skills/writing-adrs/SKILL.md — This PR modifies a file under a governance-protected path (skills/). The PR links to issue Relax the ADR writing skill’s 100-line hard limit #7283 and its description explains the rationale (dropping the ADR skill's numeric line-count hard stops), so sufficient context exists. Human approval is still always required for protected-path changes, regardless of context.
    Remediation: A human maintainer must review and approve this change to skills/writing-adrs/SKILL.md before merge; no code fix is needed.

Info

  • [scope-authorization-match] skills/writing-adrs/SKILL.md — The diff removes exactly the two hard-stop bullets named in issue Relax the ADR writing skill’s 100-line hard limit #7283 (the "Total ADR length: aim for under 80 lines" bullet and the "ADR is over 100 lines" red flag), with no additional edits. Scope matches the linked issue's authorization precisely — no scope creep. Verified: no other doc in the repository duplicates the removed line-count guidance, so no staleness was introduced elsewhere.

@fullsend-ai-review fullsend-ai-review Bot added the requires-manual-review Review requires human judgment label Sep 14, 2026

@waynesun09 waynesun09 left a comment

Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Small, low-risk docs-only change; diff matches the PR description and CI is green.

@waynesun09
waynesun09 added this pull request to the merge queue Sep 14, 2026
Merged via the queue into main with commit f0edb67 Sep 14, 2026
65 of 66 checks passed
@waynesun09
waynesun09 deleted the agent/7283-relax-adr-line-limit branch September 14, 2026 17:48
@fullsend-ai-retro

fullsend-ai-retro Bot commented Sep 14, 2026 •

Copy link
Copy Markdown

🤖 Finished Retro · ✅ Success · Started 5:50 PM UTC · Completed 5:58 PM UTC

Commit: 902a609 · View workflow run →

Runtime: claude · Model: sonnet → claude-sonnet-5 · Effort: high · Cost: $1.89

@fullsend-ai-retro

Copy link
Copy Markdown

PR #7285 (docs(#7283): drop ADR skill line-count hard stops) is a clean, low-risk run of the full pipeline: triage → code → review → human sign-off → merge, with no fix cycle needed. Triage ($0.28) correctly scoped the change to two bullets in skills/writing-adrs/SKILL.md; the code agent ($1.55) implemented exactly that diff; the review agent ($1.46) correctly flagged a Medium [protected-path] finding requiring mandatory human approval for edits under skills/ and an Info note confirming scope matched the issue with no creep. Two maintainers (ralphbean, waynesun09) approved without disagreement, one noting it was a 'small, low-risk docs-only change.' Agent definitions/skills for this run resolved from fullsend-ai/agents@main (pinned 17318c2785d8). One early review dispatch (run 34853864334) was auto-cancelled ~90s in, immediately superseded by a successful retry — routine concurrency dedup, not a failure.

One evidence note for an existing issue: the code agent self-disclosed in the PR body that make lint/pre-commit could not run as a whole due to an HTTP 403 fetching remote hook repos in its sandbox (it compensated by running individual hooks manually, which passed). The review agent's findings did not verify or reference this disclosed gap. This corroborates #2108 ('Review agent should flag when required checks couldn't run in sandbox'), which already tracks the same failure mode from a prior PR (#1579) — no new issue filed, just noting the recurrence.

The mandatory-human-approval outcome for this skills/-path edit is consistent with established policy from #1659 and #3438 (governance/process docs should not be agent-approved regardless of triviality), so no change is proposed there — the gate functioned as designed and agent/human judgment matched exactly.

One new proposal: while tracing this run, discovering the agents-repo and stage dispatches required extra investigation because the finding-agent-runs/github skill recipes point at per-stage workflow files (triage.yml, code.yml, review.yml, retro.yml) in the <org>/.fullsend dispatch repo, which have not run since ~2026-08-19/20. This pipeline generation instead dispatches directly inside fullsend-ai/fullsend via a reusable-dispatch.yml shim triggered on issues/pull_request_target events. A targeted search of fullsend-ai/agents, fullsend-ai/fullsend, and fullsend-ai/.fullsend found no existing issue tracking this staleness (two adjacent-but-different issues exist: #7011 about a stale shim header comment, and closed #5286 about a different stale doc pointer).

Proposals filed

This branch was previously deployed

1 inactive deployment
dev — 902a6096 Deployed Sep 14, 2026 by fullsend-ai-coder[bot] via behaviour #12750
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

ready-for-review Triggers review agent dispatch requires-manual-review Review requires human judgment

Projects

None yet

Development

Successfully merging this pull request may close these issues.

Relax the ADR writing skill’s 100-line hard limit

2 participants