code-research
Pass
Audited by Gen Agent Trust Hub on Nov 2, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill consists of instructional Markdown files that define a methodology for code research. It does not contain any executable scripts, command-line operations, or network requests that pose a security risk. All external references are for documentation and methodology purposes.
- [INDIRECT_PROMPT_INJECTION]: The skill's primary function is to ingest and analyze external code and documentation from GitHub repositories and local files, which represents a surface for indirect prompt injection.
- Ingestion points: External GitHub repositories such as bgauryy/octocode, JuliusBrussee/caveman, and dzhng/jevgrep, as well as local project source files (SOURCE-RESEARCH.md).
- Boundary markers: The instructions (SKILL.md) mandate cross-checking conclusions against exact source bytes, recording specific revisions, and maintaining a confidence ledger, which acts as a verification framework for data integrity.
- Capability inventory: No executable tools, subprocess calls, network exfiltration, or file-system write operations are specified in the provided skill files.
- Sanitization: The methodology relies on human-verifiable citations, version tracking, and explicit confidence labels (Direct, Supported, Inferred, Speculative) to sanitize the output from potential bias or misinformation in the source material.
Audit Metadata