gait-session-tracking

Pass

Audited by Gen Agent Trust Hub on Mar 18, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill facilitates Git-backed session tracking by executing a script located at the path stored in the GAIT_MCP_SCRIPT environment variable. This execution is performed using the $MCP_CALL utility to invoke various lifecycle tools (e.g., gait_branch, gait_record_turn).
  • [PROMPT_INJECTION]: The skill creates a surface for indirect prompt injection by aggregating and storing untrusted data from multiple actions into a consolidated audit log.
  • Ingestion points: Untrusted data enters the system through the prompt and response arguments of the gait_record_turn tool in SKILL.md.
  • Boundary markers: There are no explicit delimiters or boundary markers used in the recorded content to prevent the agent from interpreting logged history as active instructions during future log reviews.
  • Capability inventory: The skill possesses command execution and repository management capabilities via the tools defined in the $GAIT_MCP_SCRIPT.
  • Sanitization: No evidence of input validation, escaping, or filtering is present for the data recorded within the session turns.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 18, 2026, 06:12 AM
Security Audit — agent-trust-hub — gait-session-tracking