The Agent Skills Directory

[SAFE]: No malicious patterns detected. The skill performs its stated function of recording ticket metadata and human feedback after a pull request merge.
[COMMAND_EXECUTION]: The skill uses standard command-line tools like gh, git, jq, and sqlite3 to gather ticket and PR information. All command executions are parameterized or regex-validated (e.g., ticket IDs) to prevent injection.
[DATA_EXFILTRATION]: No evidence of data exfiltration to external domains. Telemetry and logs are written to local files within the user's home directory or the project workspace.
[INDIRECT_PROMPT_INJECTION]: The skill handles untrusted human input for feedback fields, but correctly escapes characters when writing to storage and uses robust parsing logic (awk/jq) when reading, preventing local execution vulnerabilities. While the stored data could theoretically influence downstream agents, this is a standard risk for logging tools and is mitigated by the structured format.

compound-estimate