best-practices-researcher
Pass
Audited by Gen Agent Trust Hub on Oct 8, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill contains a vulnerability surface for indirect prompt injection.\n
- Ingestion points: Web search results from community platforms (Reddit, Hacker News, GitHub) are processed in Mode 1, 2, and 3 (SKILL.md).\n
- Boundary markers: Absent; instructions do not require the use of delimiters or warnings to ignore instructions embedded in search results.\n
- Capability inventory: The skill is authorized to write markdown files to the project root (GROUNDED_PRACTICES.md) and documentation subdirectories (docs/decisions/).\n
- Sanitization: Absent; no mention of escaping or filtering content retrieved from the web.\n- [DATA_EXFILTRATION]: The skill reads project manifest files (e.g., package.json, go.mod, Cargo.toml) and lockfiles to determine library versions. This is standard functional behavior for a version-aware research tool and does not target sensitive system credentials or private keys.\n- [SAFE]: No evidence of prompt injection, obfuscation, persistence mechanisms, or unauthorized shell command execution was detected. The instructions follow established research methodologies.
Audit Metadata