The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data such as bug reports, log dumps, and captured network traces (Phase 1, Phase 6). This presents a surface for indirect prompt injection if the data contains malicious instructions. Ingestion points: SKILL.md (Phase 1 replaying traces, Phase 2 user descriptions), scripts/hitl-loop.template.sh (user input). Boundary markers: The ## Redact section provides secret management guidelines. Capability inventory: The skill utilizes curl, CLI tools, and script execution in SKILL.md and scripts/hitl-loop.template.sh. Sanitization: Manual redaction is instructed in SKILL.md.
[DYNAMIC_EXECUTION]: The skill instructs the agent to create 'throwaway harnesses' and provides a shell script template (scripts/hitl-loop.template.sh) intended to be modified and executed at runtime.
[COMMAND_EXECUTION]: The instructions explicitly direct the agent to use command-line tools, HTTP clients, and bash scripts for reproducing and diagnosing bugs.

diagnose