Use when execution quality depends on better thinking before action, especially for plan creation, architecture choices, non-trivial coding, refactors, task...
Security Analysis
high confidenceThis is an instruction-only meta-skill that guides the agent to run a short Socratic self-Q&A before planning or delegating; its instructions, files, and requirements are coherent and proportionate to that purpose.
The name/description (Socratic self-Q&A to improve planning and delegation) matches the content: an instruction-only skill with protocol and examples. It requires no binaries, env vars, or installs — appropriate for a behavioral guideline skill.
SKILL.md confines the agent to asking specific short questions, producing defined sections, and delegating with an explicit contract. It references only the included reference files and does not instruct reading unrelated system files, contacting external endpoints, or accessing credentials.
No install spec and no code files are present; nothing is downloaded or written to disk. This is the lowest-risk install model and matches the skill's stated role as a procedural guideline.
The skill requests no environment variables, credentials, or config paths. That is proportionate for a meta-cognitive instruction-only skill.
always:false (not persistent). The agents/openai.yaml sets allow_implicit_invocation: true, which allows the agent to call the skill autonomously — this is platform-normal for skills but worth noting because it means the skill can be invoked without an explicit user prompt.
Guidance
This skill is coherent and low-risk: it's an instruction-only helper to make the agent 'think before acting' and doesn't request secrets or install code. Before enabling, consider (1) the skill source is unknown and has no homepage — enable only if you trust the registry owner, (2) implicit invocation is allowed so the agent may call this skill autonomously; if you prefer explicit control, adjust your agent/skill invocation policies, and (3) when delegating to child agents, avoid passing secrets or broad credentials in the Delegation Contract. If unsure, review the included references/protocol.md and examples.md and test the skill on safe, non-sensitive tasks first.
Latest Release
v0.1.0
Initial release of the Socrates skill. - Introduces a concise Socratic self-Q&A protocol to clarify goals, highlight unknowns, and reduce risk before action. - Default flow includes one round of targeted questions; opens a second round only if unresolved ambiguity or risk remains. - Supports task planning, architecture choice, non-trivial coding, refactors, decomposition, and work delegation to sub-agents. - Includes clear rules for when to skip, compress, or run the full Socratic pass. - Defines specific delegation handoff requirements and mini-Socratic pass for sub-agents. - Provides strict guardrails to avoid unnecessary philosophy, verbosity, or redundant rounds.
More by @Hongyi3
Published by @Hongyi3 on ClawHub
