The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes a local validation script (.agents/scripts/spec-files.mjs) using the Node.js runtime. This script is used to verify the format of the generated spec.md file. The execution is restricted to a specific script within the project's internal directory and uses sanitized file paths.
[INDIRECT_PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface as it ingests data from external sources to generate technical specifications.
Ingestion points: Untrusted data is retrieved from dialogue history and existing .agents/cooking/*/goal.md files within the repository.
Boundary markers: The skill employs a strict template (spec-template.md) to delimit generated content sections.
Capability inventory: The agent has the capability to execute a local Node.js script for file parsing and validation.
Sanitization: The feature identifier used in file paths is forced into a kebab-case format, and the validation script enforces strict ASCII requirements for paths defined within the specification files.

to-spec