The Agent Skills Directory

[SAFE]: The skill provides a structured framework for recording and analyzing AI agent failures within a repository's local environment. It does not perform any network operations, remote code execution, or privilege escalation. All proposed changes to the repository's configuration (such as rules in AGENTS.md or CI hooks) are designed to be reviewed and confirmed by a human operator, which is an effective mitigation against automated injection or misuse.
[COMMAND_EXECUTION]: The skill instructions include standard shell commands like grep and ls for pattern matching and indexing within the local docs/reflection-log/ directory. These commands are used for legitimate data processing tasks and do not involve untrusted input or risky execution patterns.

agent-rules