save-session-context

Pass

Audited by Gen Agent Trust Hub on Jun 26, 2026

Risk Level: SAFE
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes standard, low-risk shell commands to retrieve repository status, including git status, git log, and git branch.
  • [DATA_EXPOSURE]: The skill reads local environment metadata (current branch, commit hash, and file status) and writes it to a temporary file at .tmp/current/session/context.md. No sensitive data or credentials are accessed, and no network exfiltration was detected.
  • [PROMPT_INJECTION]: There is a potential surface for indirect prompt injection as the skill ingests untrusted data from git logs (commit messages) and file names to populate the session context. While this data is written to a markdown file, a downstream agent reading this file could potentially be influenced by instructions embedded in the repository's commit history or file structure.
  • Ingestion points: git status, git log -1 (SKILL.md)
  • Boundary markers: Uses markdown headers and blockquotes to structure the output file, though it lacks explicit 'ignore instructions' warnings for the interpolated git data.
  • Capability inventory: File system write access and execution of local git commands (SKILL.md)
  • Sanitization: No explicit sanitization or escaping of the commit messages or file names is performed before they are written to the context file.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 26, 2026, 01:03 AM
Security Audit — agent-trust-hub — save-session-context