The Agent Skills Directory

[SAFE]: No malicious patterns or behaviors were detected in the skill. The logic is consistent with its stated purpose as a developer productivity tool.
[PROMPT_INJECTION]: The skill does not contain instructions that attempt to bypass safety filters or override core agent behavior. It uses structured phases and explicit state management to maintain a predictable execution flow.
[DATA_EXFILTRATION]: No unauthorized data exfiltration or network operations were identified. The skill utilizes local file system operations (writing to the .omc/ directory) and internal state management, which are standard for development-oriented agents.
[REMOTE_CODE_EXECUTION]: There is no evidence of remote script downloading or execution. The skill interacts with other agent capabilities through the official Skill() and Agent() interfaces within the platform's execution environment.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest data from the local codebase (e.g., during the 'trace' phase). It proactively mitigates the risk of indirect prompt injection by instructing the agent to treat this data as non-executable context using specific <trace-context> delimiters and an explicit 'untrusted data guard' policy.
[CREDENTIALS_UNSAFE]: No hardcoded credentials, API keys, or secrets were found in the skill documentation or execution logic.
[DYNAMIC_EXECUTION]: The skill does not perform runtime code generation or execution of untrusted strings. It relies on pre-defined orchestration steps and calls to other authorized agent skills.

deep-dive