Skill Claude CodeCodex
Use when building or maintaining a markdown routing surface that maps natural-language triggers to actions and the agent is expected to both read and write the index.
2 6d ago A 34 tokens
Skill Claude CodeCodex
Use for complex or error-compounding work — deep research, multi-source analysis, refactors, audits, migrations, or tasks where hidden state, consequential risk, delegation, repeated execution, or deferred obligations can make mistakes expensive. Reach the natural result or evidence path through the smallest safe…
2 6d ago A 101 tokens
Skill Claude CodeCodex
Structured recursive decomposition of complex, open-ended questions into a persistent map with scored branches, evidence tracking, and contradiction handling. Use when multiple plausible answers or approaches must be compared, especially for decisions under uncertainty, strategy, research, prioritization, complex…
2 6d ago A 88 tokens
Skill Claude CodeCodex
Use when a run, review, experiment, implementation, agent workflow, or artifact has failed or produced a suspect result and the user wants root cause analysis, causal trace, hypothesis challenge, repair planning, or recurrence prevention. Especially useful for contradictory artifacts, failed final gates, regressions…
2 6d ago A 78 tokens
Skill Claude CodeCodex
Use when drafting a /goal for Codex or OpenCode autonomous work, especially when the goal must be concise but self-contained, preserve important context, reference supporting documents, and steer an agent without relying on conversation history.
2 6d ago A 50 tokens
Skill Claude CodeCodex
Use when editing prompts, AGENTS.md files, role instructions, shared policies, operational principles, or other agent guidance. Revise the wording so it matches the document's job, enables the intended decisions, corrects mistaken working models, and preserves hard boundaries without procedural overfit.
2 6d ago A 60 tokens
Skill Claude CodeCodex
Use when walking through a research paper, technical report, algorithm, system design, workflow, incident, agent run, or other concrete process where the user needs an evidence-grounded explanation, causal trace, process simulation, critique, or next inspection plan. Especially useful when summarizing would be too…
2 6d ago A 94 tokens
Skill Claude CodeCodex
Use for open-ended design, strategy, architecture, research, planning, or problem-solving tasks where premature linearization would hide important structure. Start from the whole shape, zoom into concrete details, zoom out when resistance appears, revise the frame, and zoom back in.
2 6d ago A 60 tokens
Skill Claude CodeCodex
Use when the user requests a current-state handoff or when no canonical current-state source can cheaply and safely support the next decision. Default to an inline, transient delta; create or update a durable state artifact only when an ownership or context-loss boundary makes persistence necessary.
2 6d ago A 57 tokens