The Agent Skills Directory

[SAFE]: The skill consists solely of markdown instructions for the AI agent to follow during a code review session. It does not include any scripts, binary files, or external dependencies.
[COMMAND_EXECUTION]: The skill explicitly disables tool and model invocation via the disable-model-invocation: true flag in the frontmatter, ensuring that the agent cannot execute shell commands or use external tools while this skill is active.
[DATA_EXFILTRATION]: No network operations (such as curl, wget, or fetch) or hardcoded credentials were detected. The skill maintains a local 'Silent Assessment' which is an internal reasoning technique for the agent and does not involve sending data to external servers.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze untrusted data (user-provided code, diffs, or file paths). While this creates an attack surface for indirect prompt injection, the lack of any executable capabilities or network access effectively mitigates the risk of exploitation.
[PROMPT_INJECTION]: The instructions do not contain attempts to override system safety guidelines, bypass content filters, or extract system prompts. The guidance is focused on maintaining a specific pedagogical persona ('Senior pairing partner').

socratic-review