The Agent Skills Directory

[COMMAND_EXECUTION]: The skill performs dynamic execution of a Python script located within the workspace based on a file search.
Evidence: SKILL.md contains logic to find . -path "*/forge_agent/cost_scan.py" and subsequently execute python <path-to-cost_scan.py>.
Context: This pattern allows for the execution of arbitrary code if a file matching that path is present in the repository being audited.
[PROMPT_INJECTION]: The skill processes a large number of infrastructure configuration files, creating a surface for indirect prompt injection.
Ingestion points: The skill reads multiple file types across SKILL.md including .tf, Pulumi.yaml, fly.toml, render.yaml, wrangler.toml, vercel.json, netlify.toml, railway.toml, and docker-compose.yml.
Boundary markers: No delimiters or "ignore" instructions are provided to the agent before it reads these files.
Capability inventory: The skill is granted access to the Bash, Write, Edit, WebFetch, and WebSearch tools in the frontmatter of SKILL.md.
Sanitization: No input validation or content sanitization is performed on the data read from configuration files.
[EXTERNAL_DOWNLOADS]: The skill references and uses external cost analysis services and CLIs.
Evidence: SKILL.md documents dependencies on the infracost CLI and its associated API for cost auditing, and the skill includes WebFetch and WebSearch in its allowed-tools configuration.

forge-cost