The Agent Skills Directory

[SAFE]: The skill includes a dedicated guidance file (references/cribl-apps-guidance.md) that enforces security best practices for the Cribl platform, such as mandating the use of encrypted KV stores for secrets, pinning dependency versions, and restricting file operations to the local application directory boundary.\n- [COMMAND_EXECUTION]: The skill is instructed to execute npm run lint within the local environment to ensure code quality during the implementation workflow. This is a standard development practice and is considered a safe local command in this context.\n- [PROMPT_INJECTION]: The skill processes implementation instructions directly from APP_BRIEF.md, which creates a surface for indirect prompt injection if the brief's content is externally controlled.\n
Ingestion points: The skill reads APP_BRIEF.md in SKILL.md to identify implementation tasks.\n
Boundary markers: The instructions lack explicit delimiters or warnings to ignore instructions potentially embedded within the brief content.\n
Capability inventory: Across SKILL.md, the agent has capabilities for file system modification (creating/updating files) and local shell command execution (npm run lint).\n
Sanitization: No sanitization or filtering is performed on the content of the brief before the agent processes it as a source of truth.

app-implement