The Agent Skills Directory

[SAFE]: The skill is entirely composed of markdown instructions, rules, and templates. It does not contain executable code, script files, or external dependencies.
[PROMPT_INJECTION]: No evidence of malicious prompt injection, jailbreak attempts, or instructions to bypass safety filters was found. The skill includes an explicit ethics module (rules/ethics.md) that instructs the agent to refuse manipulative requests and prioritize ethical guidelines over user requests for deception.
[DATA_EXFILTRATION]: The skill does not access sensitive local files, environment variables, or hardcoded credentials. It does not perform any network operations or include URLs for data exfiltration.
[REMOTE_CODE_EXECUTION]: There are no patterns involving the download or execution of remote scripts or packages. The skill is explicitly described as advisory and does not produce or execute product code.
[DYNAMIC_CONTEXT_INJECTION]: The skill does not use dynamic shell execution placeholders (exclamation mark followed by backticks) in its markdown content.
[INDIRECT_PROMPT_INJECTION]: While the skill ingests user input to generate advisory plans, it lacks the capabilities (file-write, network, command execution) to be exploited. It uses structured templates and a predefined workflow to constrain its output, reducing the surface for injection attacks.

behavioral-design