The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests and analyzes untrusted data from external sources, creating a potential surface for indirect prompt injection attacks.
Ingestion points: Processes CI logs, PR reviews, command output, and web content as evidence during retrospectives.
Boundary markers: Includes a dedicated 'Retrieved Evidence Boundary' section in SKILL.md instructing the agent to treat external text as data rather than instructions.
Capability inventory: Has permissions to read and write files within the .tigerkit/ directory and execute local commands like git and rg (ripgrep) for repository analysis.
Sanitization: Implements a strict anonymization protocol in SKILL.md requiring the removal of PII and secrets before saving reports, which is verified by subsequent searches.

tk-retro