docs(autonomy-readiness): tighten weak language in skill

[ralphbean]

Ran skillsaw (evaluating it per #188) against the repo's skills.
Replaces vague/hedging phrasing with direct instructions: "handle
silent approvals correctly" -> "classify silent approvals as
inconclusive", and "consider adding new agent skills" -> "propose
new agent skills ... when a human reviewer consistently catches a
class of issue".

Related to #188

Assisted-by: Claude Opus 4.6 noreply@anthropic.com
Signed-off-by: Ralph Bean rbean@redhat.com

---

_{Stack created with GitHub Stacks CLI • Give Feedback 💬}

[qodo-code-review]

PR Summary by Qodo

Tighten autonomy-readiness skill guidance for silent approvals and new skills

📝 Documentation 🕐 Less than 5 minutes

AI Description

• Clarify that silent human approvals make gap deltas inconclusive, not “no gaps”.
• Replace hedging language with direct guidance on proposing new repo-level skills.

Diagram

graph TD
  A["Skill doc: Autonomy Readiness"] --> B["Evaluator guidance"] --> C["Classification rules"] --> D["Report outcomes"]
  A --> E["Suggest repo skills"]
  E --> F[".claude/ skills"] --> G["Review agents"]

Loading

High-Level Assessment

The PR’s approach (tightening wording in-place) is optimal for a guidance document: it preserves intent while removing ambiguity. Considered alternatives like adding a new section or examples, but they’d increase length without addressing the core issue (hedging language) as directly.

Files changed (1) +2 / -2

Documentation (1) +2 / -2

SKILL.mdMake silent approvals explicitly “inconclusive” and strengthen skill-addition guidance +2/-2

Make silent approvals explicitly “inconclusive” and strengthen skill-addition guidance

• Renames the silent-approval heading to explicitly require an “inconclusive” classification when humans approve without comments. Rewords the guidance on adding repo-level agent skills to be directive: propose new skills when humans repeatedly catch a missed issue class.

skills/autonomy-readiness/SKILL.md

[fullsend-ai-review]

🤖 Finished Review · ✅ Success · Started 5:00 PM UTC · Completed 5:09 PM UTC

Commit: 270d0b5 · View workflow run →

[qodo-code-review]

Code Review by Qodo

🐞 Bugs (0) 📘 Rule violations (0) 📜 Skill insights (1)

1. Protected skills/ file modified 📜 Skill insight § Compliance

Description

This PR changes skills/autonomy-readiness/SKILL.md, which is a protected governance/infrastructure
path and must not be auto-approved. A human review is required before merge to meet the
protected-path compliance requirement.

Code

skills/autonomy-readiness/SKILL.md[37]

+**Classify silent approvals as inconclusive:**

Relevance

●●● Strong

Repo enforces protected-path checks; protected-path enforcement changes were accepted previously
(human review required).

PR-#569

ⓘ Recommendations generated based on similar findings in past PRs

Evidence

The compliance rule requires raising a finding whenever a PR modifies files under protected paths
(including skills/). The diff shows modifications in skills/autonomy-readiness/SKILL.md,
triggering the requirement for human review (no auto-approval).

skills/autonomy-readiness/SKILL.md[37-39]
Skill: pr-review

Agent prompt

The issue below was found during a code review. Follow the provided context and guidance below and implement a solution

## Issue description
The PR modifies a protected path (`skills/`), which must not be auto-approved and requires explicit human review/approval before merging.

## Issue Context
`skills/` is explicitly listed as a protected governance/infrastructure path.

## Fix Focus Areas
- skills/autonomy-readiness/SKILL.md[37-37]

ⓘ Copy this prompt and use it to remediate the issue with your preferred AI generation tools

---

Context

✅ Compliance rules (platform): 55 rules

✅ Skills: 4 invoked

  code-review

  code-implementation

  pr-review

  docs-review

Tip of the day

💡 Did you know, you can group findings by type and pick your Finding display, from Minimal to Full

More tips ↗ | Customize Qodo ↗ | Qodo docs ↗

[fullsend-ai-review]

Review

Findings

Medium

• [protected-path] skills/autonomy-readiness/SKILL.md — This PR modifies a file under a protected path (skills/). The PR links to issue Evaluate skillsaw for linting agent instructions #188 and explains the rationale (applying skillsaw findings to tighten weak language). Human approval is always required for protected-path changes, regardless of context.

---

Labels: PR modifies skill documentation in skills/ directory

[waynesun09]

Approving — no blocking issues. The qodo protected-path note is satisfied by this approval itself. My own MEDIUM (skill-proposal trigger narrowing) is a non-blocking semantic nuance, not a defect. CI green.

[fullsend-ai-retro]

🤖 Finished Retro · ✅ Success · Started 9:40 PM UTC · Completed 9:52 PM UTC

Commit: 270d0b5 · View workflow run →

[fullsend-ai-retro]

Retro: PR #747 — docs(autonomy-readiness): tighten weak language in skill

Timeline

1. 2026-08-11 16:59 UTC — Ralph Bean opens PR #747, a 2-line change to skills/autonomy-readiness/SKILL.md to tighten hedging language. Linked to #188 (skillsaw evaluation).
2. 2026-08-11 17:00–17:09 UTC — Fullsend review agent runs (run 31515278039, $1.69, 27 turns). Posts one MEDIUM finding: protected-path compliance for skills/ directory. The correctness sub-agent (Opus) concludes "changes are semantically equivalent to original text." Intent-coherence sub-agent (Sonnet) finds no issues.
3. 2026-08-13 12:54 UTC — Human reviewer waynesun09 posts a substantive MEDIUM inline comment identifying that the line-102 rewrite collapses two independent statements into one conditional, narrowing the trigger for proposing repo-level skills. Notes this creates tension with the preceding paragraph's directive to "not limit yourself to these [categories]."
4. 2026-08-13 14:28 UTC — waynesun09 approves, characterizing the finding as "a non-blocking semantic nuance, not a defect."
5. 2026-08-13 21:38 UTC — PR merged by ralphbean.

Key finding: intent-coherence sub-agent's exploration budget prevented detection of semantic narrowing

The human reviewer's core insight was that the rewording changed meaning, not just tone. The original text contained an unconditional recommendation ("consider adding new agent skills") plus a conditional emphasis ("If a human reviewer consistently catches..."). The rewrite merged these into a single conditional ("propose new agent skills... when a human reviewer consistently catches..."), which reads as the sole trigger for skill proposals — contradicting the paragraph above.

The intent-coherence sub-agent is the natural owner of this class of issue (intent alignment, scope changes), but its exploration budget classifies diffs under 20 lines as trivial: 5 tool calls max, read only the diff and linked issue, then STOP. For this 4-line diff, the sub-agent would not have read the surrounding paragraphs of SKILL.md or detected the intra-document tension. The non-trivial path (20+ lines or structural changes) reads CLAUDE.md, AGENTS.md, and ADRs — but a small wording change to a skill file doesn't trigger it.

The correctness sub-agent (Opus, 15-call budget) had more room to explore but explicitly concluded the changes were "semantically equivalent" — either it also didn't read enough surrounding context, or it failed to recognize the scope narrowing.

Evidence for existing issues

• #427 (docs-only PR autonomy tracking): This PR provides a data point where the review agent missed a semantic issue that a human reviewer caught on a docs-only skill file change. The pattern matches Track autonomy evidence: review agent competently handles docs-only PRs #427's observation that "humans contributed domain-specific semantic judgment" — the agent was strong on procedural checks (protected-path) but missed the semantic narrowing.
• #744 (prompt-engineering heuristics for agent definition files): This PR shows the same gap extends to skill files (skills/*/SKILL.md), not just .fullsend/agents/*.md files. Both file types govern agent behavior and need elevated review scrutiny for small changes.

Proposals filed

• Intent-coherence sub-agent should treat changes to behavior-governing files as non-trivial regardless of diff size (in fullsend-ai/agents)