The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze untrusted external data, specifically source code, git diffs, and debugger outputs. This represents an indirect prompt injection surface where instructions embedded in the analyzed code could attempt to influence the agent's logic. However, the skill's primary purpose is technical explanation, and it does not possess dangerous capabilities like network exfiltration or persistent file modification.
Ingestion points: Reads code files, git diffs, debugger state, and the outputs of the how and why skills.
Boundary markers: The instructions emphasize maintaining a strict separation between the skill's stylistic directives and the content being explained (e.g., maintaining confidence hedges from the why tool).
Capability inventory: Invokes internal skills (how, why, unslop), utilizes an image-generation tool for diagrams, and accesses local development artifacts (code/debuggers).
Sanitization: No explicit sanitization of code comments or strings is defined in the instructions.
[COMMAND_EXECUTION]: The instructions direct the agent to utilize a debugger when necessary to explain system behavior. This involves interacting with the execution state of the project being analyzed, which is standard for developer-oriented assistant tools.

teach