reposit
Verdict safe, but 3 findings (3 medium), mainly unauthorized tool use, data exfiltration, prompt injection. Most recent scan 2026-02-07.
This skill presents MEDIUM security risks primarily related to autonomous tool execution, automatic data sharing to external services, and transitive trust in community-provided content. The skill is designed to automatically search external knowledge bases, share problem-solution data, and vote on content without consistent user confirmation. The REPOSIT_AUTO_SHARE=true configuration option enables automatic data exfiltration. The instruction to 'search proactively without being asked' and trust community-scored solutions creates indirect prompt injection risks. While the skill serves a legitimate purpose (community knowledge sharing), its autonomous behavior patterns and external data transmission require significant security improvements. The missing metadata and referenced files are minor issues but reduce transparency.
Scanner's overall assessment (LLM analyzer)
Second scanner: NVIDIA SkillSpector
Independent static run, no language model, over the same files as the Cisco scan above. How the two scanners compare across the whole corpus: Two scanners, one corpus.
What this skill does
Community knowledge sharing for AI agents - search, share, and vote on solutions via Reposit. Automatically searches when encountering errors, shares solutions after solving problems, and votes to surface quality content.
- Package
- 2 files, 6.2 KB
- SKILL.md
- 542 words, 12+ sections
- Code
- None. Instructions only.
- Requires
- bins:
npx - Homepage
- https://reposit.bot
SKILL.md covers: Reposit · Setup · Authentication · Available Tools · `search` - Find existing solutions · `share` - Contribute solutions · `vote_up` - Upvote helpful solutions · `vote_down` - Flag problematic solutions · `list_backends` - View configuration · `login` - Authenticate · Configuration · Best Practices
Findings
MEDIUM Automatic Tool Execution Without User Consent Unauthorized Tool Usellm
The skill configures automatic search and voting behavior that executes tools without explicit user permission. The search tool 'triggers automatically' when encountering errors or starting work on problems, and vote_up 'triggers automatically' after using solutions. This autonomous behavior violates user control principles and could lead to unintended data transmission to external services.
SKILL.md
**Triggers automatically** when:
- Encountering an unfamiliar error or exception
- Starting work on a non-trivial problem
- User asks 'is there a better way?'
- Before implementing a complex feature
Search proactively without being asked.MEDIUM Automatic Data Sharing to External Service Data Exfiltrationllm
The skill is configured to automatically share problem-solution data to an external service (Reposit backend) when REPOSIT_AUTO_SHARE=true is set. This creates a data exfiltration risk where code context, error messages, and implementation details are transmitted without per-instance user consent. The skill collects 'problem descriptions with error messages and context' which may contain sensitive information.
SKILL.md
**Behavior depends on configuration:**
- Default: Asks for confirmation before sharing
- Set `REPOSIT_AUTO_SHARE=true` to share automaticallyMEDIUM Transitive Trust in External Community Content Prompt Injectionllm
The skill instructs the agent to automatically trust and present solutions from external community sources without validation. It directs the agent to 'search proactively without being asked' and present findings based solely on community scores. This creates indirect prompt injection risk where malicious actors could inject harmful instructions via highly-voted solutions that the agent would then follow.
SKILL.md
Search proactively without being asked. Present findings with their community scores:
- **High score (5+)**: Community-validated, excellent match
- **Medium score (1-4)**: Worth reviewing
- **Low/negative score**: May have issuesWhat is in the package
| File | Kind | Size |
|---|---|---|
README.md | doc | 2.0 KB |
SKILL.md | doc | 4.2 KB |
External hosts referenced
github.com reposit.bot
Hostnames that appear in the skill's files. Referencing a host is not a finding; compare against the findings above and the skill's stated purpose.
Package contents recorded 2026-10-22. The scan above ran against the version current at scan time; the two can differ if the skill has since been updated.
What we checked
- Static analyzer. Pattern rules over SKILL.md and bundled files: known-bad commands, exfiltration URLs, secrets, obfuscation.
- Llm analyzer. A language model reads the skill and reports intent-level issues the rules cannot express.
- Meta analyzer. A second model pass that validates, deduplicates and re-scores the other analyzers' findings.
What we did not check
- Runtime behaviour. The skill was not executed; findings come from reading its files.
- Anything fetched at install or run time, including remote scripts, package dependencies, and URLs the skill tells the agent to open.
- Content the analyzers cannot read: images, binaries, and some non-English text.
- Known scanner bypass classes. See the methodology and limitations page.
Sources
- Repository: https://clawhub.ai/skills/reposit
- Raw scan result: reposit-collective-intelligence-for-ai-agents-scan.json
- Badge for your README:
- Request a re-scan