Name: learn
Rating: 70.39999999999999 (1 reviews)
Author: tdg-ninja

learn

Update the project's long-term memory after a merge to main. Reads the merged diff and the current memory, then writes Expert shards, discovered invariants, candidate lints, and AGENTS.md pointers — all on a reviewable learn/<sha> PR. Use post-merge (the harness invokes it automatically) or with --rebuild to regenerate memory from scratch. Triggers - learn, expert-update, update memory, update expert, post-merge memory, self-improve (project)

Quality

88%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

Securityby

Passed

No findings from the security scan

The canonical home for this skill is learn in tdg-ninja/context-specs-factory-ai

Quality

Content

85%Weight 40%Scale 1-5

Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.

A well-engineered instruction skill: a clear sequenced flow with genuine validation checkpoints and a clean one-level-deep reference structure whose files all exist. The body is dense and project-specific; its only costs are triple-statement of the headline rules and a few mechanics (worktree setup, PR creation) left implicit to the memory loop.

Suggestions

Collapse the ground-truth rule to one authoritative statement (e.g., P1) and have the callout and 'Hard nevers' reference it, cutting ~10 lines of repetition; same for the triple-stated never-auto-merge rule.

Add a short concrete command block for the worktree/branch/PR mechanics (create learn/<sha> off origin/main, push, open PR) so the skill is self-sufficient when run outside the memory loop.

Dimension	Reasoning	Score
Conciseness	The body is dense and assumes competence — no known-concept explanations — but core rules are stated three times (the ground-truth callout + P1 + 'Never write memory for uncommitted or planned work'; auto-merge in the intro + P8 + 'Never auto-merge'), which is trimmable redundancy rather than the lean ideal.	4 / 5
Actionability	Concrete guidance throughout — exact paths ('.claude/skills/expert/references/*.md', 'scripts/lints/', 'scripts/local-checks.sh'), the invocation command 'claude -p "/learn --since <sha> --sha <sha>"', the 2/3 threshold, and a real validation script — but worktree/branch/PR mechanics are delegated to the memory loop rather than specified, leaving minor gaps.	4 / 5
Workflow Clarity	Steps 0–6 are clearly sequenced with explicit validation checkpoints: idempotency pre-check via 'git ls-remote origin learn/<sha>', 'The drafted lint MUST pass against the just-merged code before you include it — run it; if it fails on current main it's wrong', and Step 5's re-run of lints plus 'scripts/check-agents-md.sh' — with feedback loops (extend-not-vote on P7 edits, discard sub-threshold votes).	5 / 5
Progressive Disclosure	A clear overview with five well-signaled, one-level-deep references — each listed with its purpose under 'How to run this skill' and cited inline at the relevant philosophy item — plus the executable 'scripts/check-agents-md.sh'; all referenced files exist in the bundle and detail is appropriately split out.	5 / 5
	Total	18 / 20 Passed

Description

88%Weight 40%Scale 1-5

Based on the skill's description, can an agent find and select it at the right time? Clear, specific descriptions lead to better discovery.

A strong description: concrete, comprehensive actions with explicit when-to-use guidance and an explicit trigger list. The only soft spots are the telegraphic 'Triggers -' keyword list (rather than natural-phrase sentence form) and slightly generic words like 'learn' that add minor conflict risk.

Dimension	Reasoning	Score
Specificity	The description lists multiple concrete actions — 'Reads the merged diff and the current memory, then writes Expert shards, discovered invariants, candidate lints, and AGENTS.md pointers — all on a reviewable learn/<sha> PR' — covering the full read/write/output surface comprehensively with no vague filler.	5 / 5
Completeness	It explicitly answers what ('Update the project's long-term memory after a merge to main…') and when ('Use post-merge (the harness invokes it automatically) or with --rebuild to regenerate memory from scratch' plus an explicit trigger list), both with concrete phrasing.	5 / 5
Trigger Term Quality	'Triggers - learn, expert-update, update memory, update expert, post-merge memory, self-improve (project)' gives good natural-phrase coverage, but the telegraphic keyword-list form skips natural sentence variants and includes the generic bare word 'learn'; coverage is solid but not synonym-complete.	4 / 5
Distinctiveness Conflict Risk	The post-merge memory-reconciliation niche with learn/<sha> PRs is unmistakable, but the bare trigger words 'learn' and 'update memory' carry minor overlap risk with generic learning/memory skills, so it sits at mostly-distinct rather than a fully clean niche.	4 / 5
	Total	18 / 20 Passed

Validation

93%

Warnings & errors only

Checks the skill against the spec for correct structure and formatting. All validation checks must pass before discovery and implementation can be scored.

Validation — 15 / 16 Passed

Validation for skill structure

Criteria	Description	Result
referenced_paths_exist	Referenced path issues: 5 missing	Warning

	Total	15 / 16 Passed

Repository: tdg-ninja/context-specs-claude-code
Path: skills/harness/learn/SKILL.md
Commit: 3a7a725

Reviewed: 2 hours ago

Table of Contents

Discovery Implementation Validation

Is this your skill?

If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.