The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill downloads the commitshow package from the NPM registry via npx commitshow@0.3.23. This package is a vendor-owned resource used to execute the core auditing logic.
[COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands to run audits (npx commitshow), manage output directories (mkdir -p .commitshow), and process result files using jq and cat.
[DATA_EXFILTRATION]: Information about the repository is sent to the remote API at https://api.commit.show for analysis. This data transfer is central to the skill's functionality, and the skill documentation provides transparency regarding the process's access to local files.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it parses externally generated JSON data (.commitshow/audit.json) and incorporates findings into the agent's response stream without explicit sanitization or strict boundary markers.
Ingestion points: .commitshow/audit.json (parsed and surfaced in Step 2 and Step 3).
Boundary markers: Absent; the agent is instructed to use the findings directly to generate follow-up questions and patches.
Capability inventory: The skill possesses the ability to execute shell commands, read/write files, and suggest code changes (patches).
Sanitization: No explicit sanitization or escaping of the external tool's output is performed before it is presented in the conversation.

production-audit