The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes the agent's conversation history to restate messages, which involves handling potentially untrusted data.
Ingestion points: The skill reads the 'last message' from the conversation context and refers to local project files like CONTEXT.md and CONTEXT-MAP.md (SKILL.md).
Boundary markers: There are no explicit delimiters or instructions provided to the agent to treat the restated content as untrusted data.
Capability inventory: The skill configuration explicitly includes disable-model-invocation: true, which prevents the agent from calling any tools or other models while processing this skill (SKILL.md).
Sanitization: No specific sanitization or validation logic is defined for the input content before it is restated.
[NO_CODE]: The skill consists entirely of natural language instructions and configuration metadata. It does not contain or reference any executable scripts, binaries, or package dependencies.

do-bro