The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data to generate executable code and server configurations.
Ingestion points: Stage 1 of the procedure fetches documentation from user-provided URLs and extracts text from local files (PDF, DOCX, Markdown).
Boundary markers: The skill does not specify explicit boundary markers or instruction-ignoring prompts when interpolating the extracted documentation into the agent's context.
Capability inventory: The skill possesses file write access, the ability to execute shell commands via bunx and uv, and performs network requests via curl.
Sanitization: Input is passed through specialized adapters and an internal extraction script (extract_text.py) before processing.
[COMMAND_EXECUTION]: The skill relies on shell command execution for its core functionality, including extraction and verification.
Evidence: The SKILL.md file instructs the agent to use uv run to execute the bundled extract_text.py script for document conversion. Stage 5 uses the smoke.sh script to boot the mock server and test routes using bunx or docker.
[EXTERNAL_DOWNLOADS]: The skill performs network operations to fetch external resources.
Evidence: Stage 1 uses curl to download documentation from user-supplied URLs. The skill also utilizes bunx to fetch the @dhaneshpurohit/mockstar package and pulls the ghcr.io/dhanesh/mockstar Docker image, which are vendor-controlled resources.
[DYNAMIC_EXECUTION]: The skill generates code and configurations that are executed at runtime.
Evidence: Stage 3 generates TypeScript dynamic handlers in the handlers/ directory when the --fidelity full flag is used. Stage 6 produces a Dockerfile based on an internal template. These generated files are then executed by the mockstar CLI or Docker.
[SAFE]: The skill implements strong security controls for credential management.
Evidence: Both inventory.schema.json and mockstar-mapping.md explicitly prohibit the use of inline secrets in generated configurations, instead enforcing references to environment variables ({{ env.NAME }}) or external secret files.

mockstar-mock