botlearn-examiner
v1.0.0Conducts standardized exams assessing multi-dimensional OpenClaw Agent capabilities, scores with rubrics, generates radar-chart reports, and offers improveme...
⭐ 0· 330·0 current·0 all-time
by邢怀康@calvinxhk
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
OpenClaw
Benign
high confidencePurpose & Capability
Name, description, manifest, SKILL.md, question bank, scoring, strategy, and tests all align with an examiner/test-runner skill. No unexpected environment variables, binaries, or external credentials are requested in the metadata that would be unrelated to capability assessment.
Instruction Scope
SKILL.md describes delivering questions, collecting JSON answers, applying rubrics, running automated checks ('Code correctness: Run against tests') and factual checks against the knowledge base. This is consistent with a testing tool, but implies the agent will need the ability to execute code/tests and access knowledge/search tools. Confirm the runtime provides appropriate sandboxes and that any automated code execution is isolated and consented to. The instructions do not request reading unrelated system files or secrets.
Install Mechanism
Instruction-only skill (no install spec, no downloads). Package files are present in the manifest but nothing is installed onto the host by the skill itself. This is the lowest-risk install pattern.
Credentials
Skill declares no required environment variables, credentials, or config paths. Example question formats reference external helper skills (e.g., @botlearn/google-search) only as recommended tools; these are not required by the skill metadata. No disproportionate credential requests are present.
Persistence & Privilege
Skill does not set always:true and is user-invocable. It does not request to modify other skills or system-wide settings. Autonomous invocation is allowed by default (disable-model-invocation: false) which is normal for skills; there are no additional privilege escalations requested.
Assessment
This skill appears to be what it claims: a standardized examiner that loads its question bank, delivers prompts, collects JSON answers, and scores them. Before installing, consider: (1) The examiner may run automated checks that execute code or run tests — ensure your agent/runtime will sandbox execution and will not run untrusted code against sensitive data. (2) The skill can reference external search or other helper skills for evidence collection; ensure you consent to any network/search operations those helper skills perform. (3) The report generation includes options to export or contribute anonymized benchmarks — review what will be shared before exporting. (4) Because the package source/owner are unknown, if you plan to run it in production or provide it with private examples, review the question bank and strategy files for any prompts that might cause unintended data disclosure. Overall the skill is internally consistent, but verify runtime isolation and data-sharing choices before use.Like a lobster shell, security has layers — review code before you run it.
latestvk97at1rycyjjtjj48kmdpcxkzh823ps7
License
MIT-0
Free to use, modify, and redistribute. No attribution required.