The Agent Skills Directory

[SAFE]: The skill implements a set of behavioral heuristics designed to improve engineering discipline, such as requiring baselines before changes and verifying sub-agent claims. It does not contain executable code, network requests, or sensitive file access.
[PROMPT_INJECTION]: The persistence mechanisms and persona instructions are standard for behavioral modification skills and include explicit boundaries that respect safety filters and user overrides.
[INDIRECT_PROMPT_INJECTION]: The skill explicitly instructs the agent to treat external reports and delegated work as untrusted, which serves as a defensive measure against potential indirect injections in multi-agent workflows.

fable-mode