skill-evaluator-srl
v1.0.0对 Skill 进行质量评估打分的 Skill，输出评分报告与改进建议。评估 skill、skill 评分、SRL 评估、skill 质量、检查 skill、skill review、skill score
⭐ 1· 50·0 current·0 all-time
by@vuact
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
OpenClaw
Benign
medium confidencePurpose & Capability
Name/description claim to evaluate skills using the SRL framework; SKILL.md, references, README and a Python analyzer script are present and implement that purpose. The presence of scripts/srl_analyzer.py (weights + report generation) and multiple reference docs is appropriate for a skill-evaluator.
Instruction Scope
SKILL.md instructs the AI to read the target Skill's SKILL.md, scripts/, README, references/ and _meta.json — which is correct for a reviewer. It also documents and intentionally treats the self-evaluation case (evaluating skill-evaluator-srl itself) as a downgraded scenario. Be aware: the skill will read project files under a 'skills' directory when asked to 'evaluate all skills', so running it will give the tool access to repository files in scope; this is expected but worth noting.
Install Mechanism
No install spec; the skill is instruction-heavy and includes a Python script that uses only the standard library. No downloads, package installs, or external installers are used in the bundle.
Credentials
The skill declares no required environment variables, no primary credential, and no config paths. Its runtime instructions ask to read local files in the target skill directory only; no credentials or unrelated env access are requested.
Persistence & Privilege
Flags are default (always:false, agent invocation allowed). The skill does not request permanent presence nor modifies other skills or system-wide settings. It outputs reports but does not persist credentials or alter agent config.
Assessment
This skill is coherent: it contains an AI-guided review workflow plus a local Python script to compute SRL scores, and it does not request credentials or perform network installs. Two things to consider before installing/using: (1) Metadata mismatch — registry owner (kn77...) differs from _meta.json owner (wulixiaobai); this is not necessarily malicious but is unexplained and worth verifying with the publisher. (2) Self-evaluation and scope — the tool can be asked to evaluate itself or scan an entire 'skills' directory; self-assessments may inflate scores due to circular reasoning (the SKILL.md explicitly notes this), and evaluating many skills will grant the tool read access to repository files. Recommended precautions: run the tool on a copy of the repository or limit the evaluation target to specific skill directories, manually review any generated evidence citations (line references) before trusting an automated SRL score, and verify the author/source if you require provenance guarantees.Like a lobster shell, security has layers — review code before you run it.
latestvk979703v5nf27xxx20hrdstend84e215
License
MIT-0
Free to use, modify, and redistribute. No attribution required.