project-related-context

Pass

Audited by Gen Agent Trust Hub on Jul 13, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill performs shallow clones of external Git repositories provided as arguments to the tool. This involves network operations to fetch data from remote sources.
  • [COMMAND_EXECUTION]: Executing Git commands is required to perform the repository cloning and analysis described in the execution modes.
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it reads and summarizes content from external repositories (manifests, documentation, and code) and writes those summaries into a persistent rule file (.claude/rules/baseline-project-related-context.md). This rule file is designed to be loaded automatically in every future session, meaning malicious instructions embedded in a sibling repository could persist and influence the agent's behavior.
  • Ingestion points: SKILL.md (via arguments), and any files within the analyzed local or remote sibling repositories.
  • Boundary markers: The skill uses YAML blocks to structure some data, but evidence notes under sibling headings may contain verbatim content from external sources.
  • Capability inventory: Reading local files, cloning remote Git repositories, and writing persistent configuration files to the project root.
  • Sanitization: There is no explicit requirement to sanitize or escape data retrieved from external repositories before it is written to the persistent rules directory.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 13, 2026, 04:04 PM
Security Audit — agent-trust-hub — project-related-context