narco-check

Memory integrity audit. Detects hallucinations, circular confirmations, and state poisoning. Runs automatically after 2 consecutive failures or at nightly deep dive. Uses Opus 4.6 as the auditor model.

Quality

61%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

Securityby

Critical

Do not install without reviewing

Fix and improve this skill with Tessl

tessl review fix ./examples/jason-my-claw-is-the-law-deebee-4567b4/skills/narco-check/SKILL.md

Quality

Content

77%

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

The content is highly actionable with executable commands, a clearly sequenced grounding-first audit workflow, and explicit validation/escalation checkpoints. It is slightly held back on conciseness by repetitive verbatim command lists and on progressive disclosure by referencing an audit-prompt.md that is not bundled and inlining time-sensitive details.

Suggestions

Bundle the referenced audit-prompt.md in ./references/ (or inline it) so the cat-based invocation resolves; an un-bundled referenced file is a broken one-level-deep reference.

Tighten the 5-4-3-2-1 grounding block by collapsing the repeated bash fence boilerplate, since the per-step command lists compete for the context budget without adding clarity.

Move the time-sensitive nightly trigger ('~3 AM') and model version into a clearly labeled section so version-dependent details do not degrade conciseness.

Dimension	Reasoning	Score
Conciseness	The body is mostly efficient and assumes Claude's competence (no explanatory padding about what audits or hallucinations are), but the 5-4-3-2-1 grounding block repeats long verbatim command lists that are partly filler and could be tightened, matching the mostly-efficient-but-could-be-tightened anchor.	2 / 3
Actionability	It provides fully executable, copy-paste-ready bash commands (openclaw agent invocation, date/hostname/df/docker, curl pings, exa.sh) plus a concrete grounded-fact output template, matching the fully-executable anchor.	3 / 3
Workflow Clarity	The audit is clearly sequenced — run the 5-4-3-2-1 grounding check first, then read logs, then classify into CLEAN/DEGRADED/POISONED states — with an explicit validation checkpoint ('A grounded check with fabricated output is POISONED') and a feedback/escalation loop (quarantine + notify Ludo), matching the explicit-validation anchor.	3 / 3
Progressive Disclosure	There are no references/scripts/assets bundle files present, yet the body inline-references an external audit-prompt.md via --message "$(cat .../audit-prompt.md)" that is not bundled or linked, and a ~3 AM nightly trigger is hard-coded inline rather than separated out, so structure exists but key detail lives outside a missing referenced file.	2 / 3
	Total	10 / 12 Passed

Description

45%

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

The description is specific about capabilities and third-person, but it is written in internal jargon with no natural user trigger terms and only an implied (automatic) usage condition rather than an explicit 'Use when' clause. Completeness is therefore capped at 2 and trigger-term quality scores low.

Suggestions

Add an explicit 'Use when...' clause stating when a user should invoke the skill (e.g. 'Use when you suspect memory entries are fabricated or after repeated task failures').

Replace internal jargon with natural user-facing keywords a person would actually say, such as 'hallucinations', 'false memories', 'memory checks', or 'failed tasks'.

Clarify the trigger so it reads as user-invocable guidance rather than solely an automatic background process.

Dimension	Reasoning	Score
Specificity	Lists multiple concrete actions — "Detects hallucinations, circular confirmations, and state poisoning" — naming the exact failure classes the audit catches, matching the multiple-specific-actions anchor.	3 / 3
Completeness	It clearly states what the skill does but the "when" is framed as an automatic trigger ("Runs automatically after 2 consecutive failures or at nightly deep dive") rather than an explicit "Use when..." clause for the user, so per the guidelines completeness caps at 2.	2 / 3
Trigger Term Quality	There are no natural user-facing keywords; the description leans on internal jargon ("circular confirmations", "state poisoning", "deep dive", "auditor model") rather than terms a user would actually say, matching the jargon/no-natural-keywords anchor.	1 / 3
Distinctiveness Conflict Risk	The hallucination/state-poisoning framing carves a recognizable niche, but the absence of natural trigger terms and the generic "audit" framing mean it could still overlap with other diagnostic or memory skills, matching the somewhat-specific-but-overlapping anchor.	2 / 3
	Total	8 / 12 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
frontmatter_unknown_keys	Unknown frontmatter key(s) found; consider removing or moving to metadata	Warning

	Total	15 / 16 Passed

Repository: open-gitagent/opengap
Path: examples/jason-my-claw-is-the-law-deebee-4567b4/skills/narco-check/SKILL.md
Commit: d7a8e2e

Reviewed: 4 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.