The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill operates by having the agent analyze and reason about external code diffs, architecture decisions, and design options. This creates a surface area for indirect prompt injection where instructions embedded in the ingested data could potentially influence the agent's behavior.
Ingestion points: Code diffs and external design context processed in SKILL.md and modules/tradeoff-ledger.md.
Boundary markers: None defined; the skill does not specify delimiters or warnings for the agent to ignore instructions within the code being analyzed.
Capability inventory: The skill assumes the agent possesses standard file-system capabilities (e.g., reading and writing files) to perform its mentoring and documentation tasks.
Sanitization: No explicit logic is provided to sanitize or escape instructions that might be present in the source code or design documentation provided to the agent.

assisted-mastery