The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/build_evidence_inventory.py executes local git commands using subprocess.check_output to analyze repository history and file hashes. These commands are constructed using argument lists rather than shell strings, which mitigates standard command injection risks.
[DYNAMIC_EXECUTION]: The skill utilizes subprocess at runtime to interact with the Git binary. While this involves executing system-level tools using user-provided paths, the scope is restricted to the git executable and the diagnostic audit of the local project directory.
[INDIRECT_PROMPT_INJECTION]: The skill provides a surface for indirect prompt injection due to its requirement to ingest and process untrusted external data (website source code and assets). * Ingestion points: The build_evidence_inventory.py script and the agent's inspection instructions read text and metadata from the user-specified audit directory. * Boundary markers: The skill instructions do not define explicit delimiters or 'ignore' instructions for the content being audited. * Capability inventory: The agent has the capability to read local files and execute the included inventory script, which in turn calls Git subprocesses. * Sanitization: The audit script performs text normalization for comparison purposes but does not include mechanisms to sanitize content for malicious prompt instructions embedded in the analyzed files.

audit-reference-originality