The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes untrusted data from local files (.harnessed/contract.md, .harnessed/qa-state.md) and user-supplied micro-task requests to define its verification criteria and generate summary reports. 1. Ingestion points: Untrusted data enters the agent context through configuration and state files located in the .harnessed/ directory. 2. Boundary markers: Absent; there are no explicit delimiters or instructions to ignore potential instructions embedded within the contract or state files. 3. Capability inventory: The agent is instructed to use subprocess calls for curl (smoke tests), semgrep, CodeQL, and bandit (static analysis), and it performs file-write operations to the .harnessed/ directory. 4. Sanitization: Absent; there is no evidence of escaping, validation, or filtering of external content before it is interpolated into the workflow.
[COMMAND_EXECUTION]: The skill directs the agent to execute shell commands to gather evidence of task completion, including curl for network-level smoke tests and various security-oriented static analysis tools like semgrep and bandit.

verification-gate