The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill establishes an attack surface for indirect prompt injection by instructing the agent to ingest and analyze potentially attacker-controlled data.
Ingestion points: Untrusted data enters the agent context in 'Phase 1: Symptom Capture' and 'Phase 3: Evidence Analysis', where the agent is prompted to process user-provided or environment-extracted stacktraces, log files, and event payloads.
Boundary markers: The instructions do not require the use of explicit delimiters (e.g., XML tags or triple quotes) or provide 'ignore embedded instructions' warnings when interpolating this external data into the analysis prompt.
Capability inventory: The skill employs powerful local capabilities, including executing git commands for bisection, reading arbitrary source code and log files, and running test scripts or harnesses via the command line.
Sanitization: The methodology lacks requirements for sanitizing, escaping, or validating the content of analyzed logs or stacktraces before they are interpreted by the agent.

debug-investigator