The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data which could contain malicious instructions designed to influence agent behavior.
Ingestion points: SKILL.md describes accepting pasted specs, issues, and acceptance criteria. references/fan-pathway.md mentions processing wiki_hits from the repository wiki corpus.
Capability inventory: The skill can execute shell commands (quality gates), perform source code edits, and dispatch multiple other agent skills including /press, /age, and /cure.
Boundary markers: SKILL.md implements a mandatory "Contract" step where the agent must confirm behavior, scope, and gates before implementation begins.
Sanitization: The skill uses bundled CLI tools (normalize and validate) to enforce schemas on agent-authored JSON payloads before they are processed by the host.
[COMMAND_EXECUTION]: The skill executes arbitrary shell commands defined within the target project's environment.
Evidence: SKILL.md and references/quality-gates.md instruct the agent to run "existing project commands," "relevant tests," and "quality-gate commands" (lint, type, build) to verify changes.
[EXTERNAL_DOWNLOADS]: The skill references an external registry for schema validation.
Evidence: Files such as SKILL.md and references/writer-views.md reference https://schemas.easy-cheese.dev/ for fetching or identifying payload schemas. This domain serves as a well-known service for the Easy Cheese ecosystem.

cook