dotnet
Analyzes a completed PR to extract lessons learned from agent behavior. Use after any PR with agent involvement - whether the agent failed, succeeded slowly, or succeeded quickly. Identifies patterns to reinforce or fix, and generates actionable recommendations for instruction files, skills, and documentation.
bunx add-skill dotnet/maui -s learn-from-prLoading…
Extracts lessons learned from a completed PR to improve repository documentation and agent capabilities.
| Input | Required | Source |
|---|---|---|
| PR number or Issue number | Yes | User provides (e.g., "PR #33352" or "issue 33352") |
| Session markdown | Optional | CustomAgentLogsTmp/PRState/issue-XXXXX.md or pr-XXXXX.md |
Learning Analysis - Structured markdown with:
Actionable Recommendations - Each with:
The skill is complete when you have:
pr-finalize first)# Required: Get PR info
gh pr view XXXXX --json title,body,files
gh pr diff XXXXX
# Check for session markdown
ls CustomAgentLogsTmp/PRState/issue-XXXXX.md CustomAgentLogsTmp/PRState/pr-XXXXX.md 2>/dev/null
If session markdown exists, extract:
Analyzing without session markdown:
When no session file exists, you can still learn from:
Focus on: "What would have helped an agent find this fix faster?"
Critical question: Did agent attempts target the same files as the final fix?
# Where did final fix go?
gh pr view XXXXX --json files --jq '.files[].path' | grep -v test
# If session markdown exists, compare to attempted files
| Scenario | Implication |
|---|---|
| Same files | Agent found right location |
| Different files | Major learning opportunity - document why |
If different files: Answer these questions:
Determine which scenario applies and look for the relevant patterns:
| Pattern | Indicator |
|---|---|
| Wrong file entirely | All attempts in File A, fix in File B |
| Tunnel vision | Only looked at file mentioned in error |
| Trusted issue title | Issue said "crash in X" so only looked at X |
| Pattern not generalized | Fixed one instance, missed others |
| Didn't search codebase | Never found similar code patterns |
| Missing platform knowledge | Didn't know iOS/Android/Windows specifics |
| Wrong abstraction layer | Fixed handler when problem was in core |
| Misread error message | Error pointed to symptom, not cause |
| Incomplete context | Didn't read enough surrounding code |
| Over-engineered | Complex fix when simple one existed |
| Pattern | Indicator |
|---|---|
| Correct file, wrong approach | Found right file but tried wrong fixes first |
| Needed multiple iterations | Each attempt got closer but wasn't quite right |
| Discovery was slow | Eventually found it but search was inefficient |
| Missing domain knowledge | Had to learn something that could be documented |
Key question: What would have gotten agent to the solution faster?
| Pattern | Indicator |
|---|---|
| Good search strategy | Found right file immediately |
| Understood the pattern | Recognized similar issues from past |
| Documentation helped | Existing docs pointed to solution |
| Simple, minimal fix | Didn't over-engineer |
Key question: What made this work? Should we reinforce this pattern?
# Discover where agent guidance lives
find .github/instructions -name "*.instructions.md" 2>/dev/null
find .github/skills -name "SKILL.md" 2>/dev/null
ls docs/design/ 2>/dev/null
ls .github/copilot-instructions.md 2>/dev/null
| Location | When to Add Here |
|---|---|
.github/instructions/*.instructions.md | Domain-specific AI guidance (testing patterns, platform rules) |
.github/skills/*/SKILL.md | Skill needs new step, checklist, or improved workflow |
/docs/design/*.md | Detailed architectural documentation |
.github/copilot-instructions.md | General AI workflow guidance |
| Code comments | Non-obvious code behavior |
For each recommendation, provide:
Prioritization factors:
Pattern-to-Improvement Mapping (Failures/Slow Success):
| Pattern | Likely Improvement |
|---|---|
| Wrong file entirely | Check /docs/design/ for component relationships |
| Tunnel vision | Instruction file: "Always search for pattern across codebase" |
| Missing platform knowledge | Platform-specific instruction file |
| Wrong abstraction layer | Reference /docs/design/HandlerResolution.md |
| Misread error message | Code comment explaining the real cause |
| Over-engineered | Skill enhancement: "Try simplest fix first" |
Pattern-to-Improvement Mapping (Quick Success - reinforce):
| Pattern | Improvement |
|---|---|
| Good search strategy | Document the search pattern that worked in skills |
| Documentation helped | Note which docs were valuable, ensure they stay updated |
| Recognized pattern | Add to instruction files as known pattern |
Present your analysis covering:
| Situation | Action |
|---|---|
| PR not found | Ask user to verify PR number |
| No session markdown | Analyze PR diff only, note limited context |
| No agent involvement evident | Ask user if they still want analysis |
| Can't determine failure mode | State "insufficient data" and what's missing |
PR #33352 - TraitCollectionDidChange crash on MacCatalyst
What happened:
ShellSectionRootRenderer.csPageViewController.csFailure Mode: Trusted issue title instead of searching for pattern.
Recommendation:
.github/instructions/ios-debugging.instructions.mdPR #34567 - CollectionView scroll position not preserved
What happened:
Pattern: Wrong abstraction layer - fixed handler when problem was in core.
Recommendation:
.github/architecture/handler-vs-core.mdPR #35678 - Button disabled state not updating
What happened:
Pattern: Good search strategy - recognized pattern from similar code.
Recommendation:
.github/skills/try-fix/SKILL.mdFor automated application of recommendations, use the learn-from-pr agent instead.
Use when you need to run Flow type checking, or when seeing Flow type errors in React code.
Use when you want to validate changes before committing, or when you need to check all React contribution requirements.
Use when feature flag tests fail, flags need updating, understanding @gate pragmas, debugging channel-specific test failures, or adding new flags to React.
Use when you need to check feature flag states, compare channels, or debug why a feature behaves differently across release channels.