The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to use local CLI tools such as watch-skill doctor, watch-skill setup-vision, and watch-skill list to perform environment diagnostics and remediation.
[INDIRECT_PROMPT_INJECTION]: The skill ingests potentially untrusted data from tool outputs (watch-skill doctor --json) to determine recovery steps. While the instructions emphasize following structured fixes, this remains an ingestion point for external data into the agent's context. * Ingestion points: Output from the watch-skill doctor --json command in SKILL.md. * Boundary markers: None explicitly defined; the agent processes JSON structures. * Capability inventory: The agent has access to the Bash tool to execute arbitrary commands as part of the recovery process across all scripts. * Sanitization: Instructions require the agent to apply only the fixes stated in the tool output, relying on the tool's internal logic.

recovering-from-errors