The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests and processes arbitrary code and configuration files from a source directory. These files could contain instructions designed to deceive the LLM agents during the sanitization or packaging phases. 1. Ingestion points: Project files are read from the SOURCE_PATH provided by the user. 2. Boundary markers: The orchestration logic does not specify explicit markers to separate project data from agent instructions. 3. Capability inventory: The skill possesses the ability to read/write to the filesystem in a staging directory and use the gh CLI for repository management. 4. Sanitization: The skill includes a specific opensource-sanitizer sub-agent that performs multi-category scans (Secrets, PII, Internal refs) to mitigate risks.
[COMMAND_EXECUTION]: The skill automates repository management and filesystem organization using standard shell utilities and the GitHub CLI tool. Evidence: Execution of mkdir, ls, cat, and gh CLI for creating and pushing public repositories. Safety is maintained through mandatory user approval checkpoints before any public deployment.

opensource-pipeline