andywer/agent-skills

andywer/agent-skills

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.

andywer/agent-skills

Skill Claude CodeCodex

Use for broad research and exploration, complex investigations or decisions, conflicting evidence, challenged conclusions, and reconstructing prior inquiries. Connect questions, findings, alternatives and uncertainties to guide investigation. Skip straightforward tasks and ordinary summaries.

andywer/agent-skills

Skill Claude CodeCodex

Investigate failed or suspect results, inefficient processes, or work that is not closing in on its goal despite successful individual steps. Use for causal diagnosis, challenging explanations, choosing repairs, and preventing recurrence, including reviews of AI-agent processes.

andywer/agent-skills

Skill Claude CodeCodex

Review and revise agent instructions, including prompts, AGENTS.md files, and skills, to clarify intended decisions without adding unnecessary procedure or weakening real constraints.

andywer/agent-skills

Skill Claude CodeCodex

Use for source-grounded explanations of papers, algorithms, systems or concrete processes, including mechanism walkthroughs, causal traces, simulations and critiques. Skip simple summaries and broad literature surveys.

andywer/agent-skills

Skill Claude CodeCodex

Use for handoffs, context-loss recovery, or current-state briefings that require reconciling multiple or conflicting sources. Do not use when one current source answers the request directly. Default to an inline delta; persist it only across an ownership or context-loss boundary.

andywer/agent-skills

Skill Claude CodeCodex

Write or revise handovers, concepts, implementation briefs, specifications, and READMEs when intent, commitments, assumptions, or change authority need clarification. Preserve meaning while making the reader's next decisions actionable. Not for formatting-only edits, agent instruction design, or reconstructing current…

andywer/agent-skills

Skill Claude CodeCodex

Draft or revise /goal statements at the requested level—high-level objectives, bounded task contracts, or autonomous-session charters—while keeping essential success conditions and constraints self-contained.

andywer/agent-skills

Skill Claude CodeCodex

Use for complex work where accumulated context, hidden or changing state, coordination, or consequential evidence and actions can compound errors.