technical-skill-finder
Pass
Audited by Gen Agent Trust Hub on Oct 3, 2026
Risk Level: SAFEDATA_EXFILTRATIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [DATA_EXFILTRATION]: The skill is designed to read and analyze local agent log files, which may contain sensitive coding history, project paths, and commands.
- Evidence:
SKILL.mdidentifies~/.codex/history.jsonl,~/.codex/sessions/*.jsonl, and other session logs as primary data sources. - Context: This access is central to the skill's primary function of log mining, and the skill includes a required
PRIVACY_POLICYinput to mitigate accidental exposure. - [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from external log files and repository telemetry, which could theoretically contain malicious instructions designed to influence the agent's summary or ranking output.
- Ingestion points: The agent reads
~/.codex/history.jsonl,AGENTS.md, and optional MCP-exposed resources (e.g., messaging channels). - Boundary markers: The instructions do not specify explicit delimiters (such as XML tags or triple quotes) to separate raw log content from the agent's internal processing logic.
- Capability inventory: The skill performs classification, clustering, and summarization, outputting findings directly to the user.
- Sanitization: The workflow includes high-level guardrails to skip binary files and avoid private content, but lacks robust technical sanitization of the ingested log text.
Audit Metadata