The Agent Skills Directory

[SAFE]: The skill follows established software engineering and project management practices (PDCA, AAR, KPT) for reflecting on work and solidifying lessons learned.
[DATA_EXPOSURE_AND_EXFILTRATION]: No sensitive file paths, hardcoded credentials, or network exfiltration patterns were detected. The skill references configuration files like AGENTS.md and .cursor/rules/ for the purpose of storing workflow conventions, which is standard behavior for development-oriented agents.
[REMOTE_CODE_EXECUTION]: No patterns of downloading or executing remote code from external sources were found. The skill operates entirely within the provided instructions and local file system context.
[COMMAND_EXECUTION]: The skill does not include any shell command execution or dynamic context injection markers (!command).
[PROMPT_INJECTION]: The instructions do not contain attempts to override system safety guidelines or bypass AI constraints.
[OBFUSCATION]: No obfuscated content, such as Base64-encoded strings, zero-width characters, or homoglyphs, was detected in the skill files.
[INDIRECT_PROMPT_INJECTION]: While the skill processes untrusted data (user-completed work), it includes explicit guardrails requiring user confirmation before writing any changes to long-term configuration files (AGENTS.md, CLAUDE.md, etc.), mitigating the risk of automated injection attacks.

learn-and-improve