maintain

Brain health checks: back-link enforcement, citation audit, filing validation, stale info detection, orphan pages, and benchmarks. Use when asked to check brain health, run maintenance, or audit quality.

Quality

62%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

Securityby

Low

Low-risk findings worth noting

Fix and improve this skill with Tessl

tessl review fix ./skills/maintain/SKILL.md

Quality

Content

50%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.

The body is rich in gbrain-specific, actionable detail with a strong autonomous-path workflow, but it is far too long and structurally untidy: a monolithic 528-line document with a ~100-line configuration deep-dive inline, scattered version numbers, duplicated paragraphs, and a manual path that breaks off mid-sequence. Splitting reference material into bundle files and tightening the manual path would address the main weaknesses.

Suggestions

Move the dream-cycle internals (synthesize triage gate, rescue rules, quote verify, cooldown, config knobs) into a references/dream-cycle.md and keep only the invocation patterns and quality bar in SKILL.md — this fixes both the conciseness and progressive-disclosure weaknesses.

Repair the manual path: replace the dangling "2. Check each dimension:" with a real numbered list of dimensions, and delete or relocate the orphaned "Parses `- **YYYY-MM-DD**` ..." paragraph at lines 263-264 that duplicates the Graph-population note.

Collect version-pinned guidance (v0.11.0 half-migrated install repair, v0.11.1+ autopilot job dispatch) into a single "Version-specific repairs" or deprecated section so the main workflow stays version-agnostic.

Dimension	Reasoning	Score
Conciseness	The body is noticeably verbose: a ~100-line dream-synthesize configuration deep-dive (triage gates, rescue floors, quote-verify knobs, cooldown) reads like a design doc rather than instructions, version numbers (v0.36.4.0, v0.23, v0.11.1+, v0.10.3+) are scattered rather than placed in a deprecated/old-patterns section, and lines 263-264 duplicate the timeline-format parsing note in an orphaned spot. It is above a 1 because nearly all of the content is gbrain-specific knowledge Claude does not already have.	2 / 5
Actionability	Concrete, copy-paste-ready commands are abundant ("gbrain doctor --remediate --yes --target-score 90 --max-usd 5", "gbrain dream retriage --dry-run", "nohup gbrain embed refresh > /tmp/gbrain-embed.log 2>&1 &"). Not a 5 because several dimensions (back-link enforcement, citation audit, tag consistency, open threads) give only prose direction with no example command or expected output.	4 / 5
Workflow Clarity	The autonomous path is well sequenced with validation (preview the plan, re-check score between steps, bail with what's missing), and before/after health checks are required. But the manual path's numbered list breaks off mid-step ("2. Check each dimension:") into an unnumbered stream of H3 sections with duplicated/orphaned text, so the primary multi-step process has structural gaps and implicit checkpoints.	3 / 5
Progressive Disclosure	No bundle files exist, so the entire skill is one 528-line monolith with good section headers. Content that clearly belongs in a separate reference file (dream cycle internals, half-migrated install repair) is fully inlined, and external doc pointers (docs/guides/minions-fix.md, skills/_brain-filing-rules.md) are mentioned inline rather than organized as a navigable reference layer. Above a 2 because headers give real structure; below a 4 because large extractable sections remain inline.	3 / 5
	Total	12 / 20 Passed

Description

75%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 solid description: domain-specific, third-person, with a concrete action list and an explicit Use-when clause. Its main weakness is undercoverage — the description and its when-clause expose only a fraction of the triggers and capabilities the body actually serves.

Suggestions

Extend the when-clause to name the high-frequency natural triggers the skill already knows it serves, e.g. "or when asked to run the dream cycle, synthesize conversations, or backfill the link graph".

Add one or two of the missing capability keywords (e.g. "graph/link extraction", "embedding refresh") to the action list so the description matches the body's scope.

Dimension	Reasoning	Score
Specificity	The description lists six concrete actions ("back-link enforcement, citation audit, filing validation, stale info detection, orphan pages, and benchmarks"), matching the several-specific-actions-with-minor-gaps anchor. It is not a 5 because major capabilities of the skill (dream cycle, graph population, embedding refresh, autopilot) go unmentioned, and not a 3 because coverage clearly exceeds 1-2 actions.	4 / 5
Completeness	Both parts are explicit: a concrete "what" (the action list) and a "Use when asked to check brain health, run maintenance, or audit quality" when-clause. Not a 5 because the when-clause names only three trigger scenarios while the skill's own trigger list shows many more (dream, retriage, backfill, autopilot) it should fire on.	4 / 5
Trigger Term Quality	Natural phrases users would say appear ("brain health", "run maintenance", "audit quality", plus "orphan pages" and "back-link" inlined in the action list). Not a 5 because common natural variations the skill itself lists as triggers ("check backlinks", "run dream", "synthesize my conversations", "backfill graph") are absent.	4 / 5
Distinctiveness Conflict Risk	"Brain health checks" with gbrain-specific vocabulary carves a distinct niche, but the trailing "run maintenance" or "audit quality" triggers are generic and could mildly overlap with other upkeep or code-audit skills. Mostly distinct with minor overlap risk.	4 / 5
	Total	16 / 20 Passed

Validation

81%

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 — 13 / 16 Passed

Validation for skill structure

Criteria	Description	Result
skill_md_line_count	SKILL.md is long (528 lines); consider splitting into references/ and linking	Warning
frontmatter_unknown_keys	Unknown frontmatter key(s) found; consider removing or moving to metadata	Warning
relative_links	Relative link issues: 3 missing	Warning

	Total	13 / 16 Passed

Repository: garrytan/gbrain
Path: skills/maintain/SKILL.md
Commit: e78f1c3

Reviewed: 5 days 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.