fix(#1099): resolve skillsaw CI failures

[fullsend-ai-coder]

Summary

Fixes skillsaw CI failures that were introduced when the Jira source-code overlay PR (#1075) merged and increased the token count of skills/code-implementation/SKILL.md beyond its baselined ceiling.

Three changes:

• skills/code-implementation/SKILL.md: Replace vague "properly" with specific phrasing; remove non-actionable "note that" and state the constraint directly. Resolves content-weak-language violations at lines 161 and 195.
• skills/pr-risk-assessment/SKILL.md: Add "Use when" trigger phrasing to the skill description. Resolves content-description-routing violation.
• .skillsaw-baseline.json: Regenerate baseline to capture the new context-budget ceiling (11,649 tokens, up from 11,477). Fixed violations are removed from the baseline.

Verification

skillsaw --strict
# ✓ All checks passed! (0 errors, 0 warnings, 10 suppressed by baseline)

---

Closes #1099

Post-script verification

• Branch is not main/master (agent/1099-fix-skillsaw-ci)
• Secret scan passed (gitleaks — 48511880eaea5ef01f80b69ba4f228147611db33..HEAD)
• PR body secret scan passed (gitleaks — no-git)

[fullsend-ai-review]

🤖 Review · ⚠️ Cancelled · Started 9:13 AM UTC · Ended 9:18 AM UTC

Commit: 758f741 · View workflow run →

[fullsend-ai-review]

🤖 Finished Review · ✅ Success · Started 9:21 AM UTC · Completed 9:38 AM UTC

Commit: 555cc76 · View workflow run →

Runtime: claude · Model: opus → claude-opus-4-6 · Effort: high · Cost: $3.76

[fullsend-ai-review]

Risk Assessment: moderate (2/5)

Details

Small config-only PR by a trusted bot fixing skillsaw CI baseline and SKILL.md frontmatter across 4 files (92 lines). Protected paths raise a flag but the changes are low-complexity metadata corrections with clear issue scope, no security impact, and easy rollback. Moderate git churn on code-implementation/SKILL.md contributes minor history risk.

[fullsend-ai-review]

Review

Findings

Medium

• [protected-path] skills/code-implementation/SKILL.md, skills/jira-components/SKILL.md, skills/pr-risk-assessment/SKILL.md — PR modifies 3 files under the skills/ protected path. The PR links to issue Fix skillsaw CI failures #1099 and the description explains the rationale for the changes. Human approval is always required for protected-path changes, regardless of context.

Low

• [scope-creep] skills/jira-components/SKILL.md:3 — This file is not mentioned in the linked issue Fix skillsaw CI failures #1099 or the PR body's change list. The "Use when" trigger phrasing fix is consistent with the pr-risk-assessment change but extends scope beyond what was explicitly authorized. The base-branch baseline contains no content-description-routing violation for this file, so this change does not fix any existing CI failure.

[fullsend-ai-review]

See the review comment for full details.

[fullsend-ai-retro]

🤖 Finished Retro · ✅ Success · Started 10:28 AM UTC · Completed 10:39 AM UTC

Commit: 555cc76 · View workflow run →

Runtime: claude · Model: opus → claude-opus-4-6 · Effort: high · Cost: $3.38

[fullsend-ai-retro]

Retro: PR #1101 — fix(#1099): resolve skillsaw CI failures

Workflow quality: Good. The code agent correctly fixed all three skillsaw violations that existed at the time it ran ($1.60). Human rework was needed only because PR #1085 merged ~3 hours later, introducing skills/jira-components/SKILL.md with a new content-description-routing violation. The human (Hector Martinez) pushed a second commit to fix this new violation so the PR could achieve its stated goal of making skillsaw CI green.

Timeline:

1. Aug 31 17:10 — ralphbean files issue Fix skillsaw CI failures #1099 and triggers /fs-code
2. Aug 31 17:11–17:17 — Triage agent runs ($0.83), correctly identifies violations in code-implementation/SKILL.md and pr-risk-assessment/SKILL.md
3. Aug 31 17:19–17:25 — Code agent creates PR fix(#1099): resolve skillsaw CI failures #1101 with fixes for 3 files
4. Aug 31 17:25–17:26 — First review run fails: fullsend-ai/agents@v0.38.0 tag 404 (tag hadn't been pushed to agents repo yet)
5. Aug 31 18:24 — Human approves PR (before review agent successfully runs)
6. Aug 31 20:21 — PR feat(#1073): add Jira component assignment to triage pipeline #1085 merges, introducing skills/jira-components/SKILL.md
7. Sep 01 09:11 — Branch rebased onto updated main
8. Sep 01 09:17 — Human pushes fix for jira-components content-description-routing violation
9. Sep 01 09:21–09:38 — Review agent succeeds ($3.76), posts risk assessment (moderate 2/5) and two findings
10. Sep 01 10:26 — PR merged

Review quality: The review agent posted two findings. The protected-path finding (medium) is correct and valuable — human approval is required for skills/ changes. The scope-creep finding (low) on skills/jira-components/SKILL.md is technically correct (file not in issue #1099) but contextually misleading — the file didn't exist when the issue was filed, and the human fixed it to achieve the PR's goal. See proposal below.

Existing issue corroboration:

• The first review run's tag-resolution 404 failure is covered by fullsend#6824 (CLI agent resolution should fall back to main when version tag is missing). This retro provides additional evidence: the v0.38.0 tag was pushed to the agents repo ~1m46s after the review agent tried to resolve it, causing a hard failure with no fallback.
• Issue #1091 covers the related problem of re-review diffs including unrelated main-branch commits after rebase, but from the diff-computation side rather than the scope-authorization judgment side.

Agents repo: Both code and review agents resolved from fullsend-ai/agents@main (code at commit 48511880eaea, review at commit c4879824d504).

Proposals filed

• Scope-authorization sub-agent should recognize concurrent-merge-introduced changes as goal-aligned, not scope-creep (in fullsend-ai/agents)