recipe-diagnose

Required Skills [LOAD BEFORE EXECUTION]

Before loading required skills, use wc -c to size the files and set both exec_command.max_output_tokens and the outer // @exec: {"max_output_tokens": N} budget to accommodate their contents, including the combined output of batched reads. Split oversized reads upfront; recover any truncated content before proceeding.

1. [LOAD IF NOT ACTIVE] ai-development-guide — AI development patterns
2. [LOAD IF NOT ACTIVE] coding-rules — coding standards
3. [LOAD IF NOT ACTIVE] llm-friendly-context — clear prompts, handoffs, and generated artifacts

Invocation rule: Start each named agent with spawn_agent and fork_turns="none". When invoking the same named agent again within this flow, use followup_task on its existing instance with the inputs specified for the current invocation.

Context: Diagnosis flow to identify concrete failure points and present solutions

Target problem: $ARGUMENTS

Orchestrator Definition

Execution Method:

• Investigation -> Invoke investigator agent
• Verification -> Invoke verifier agent
• Solution derivation -> Invoke solver agent

The orchestrator structures the reported problem, coordinates the three specialist stages, evaluates their results, and passes only the context needed by the next stage.

Execution Plan: Reuse the active execution plan. When the workflow has multiple dependent actions and no plan exists, create one that tracks them through final verification. Complete a plan step after verifying its result; start a dependent step after its prerequisites are satisfied.

Step 0: Problem Structuring (Before invoking investigator)

Keep the user's current symptom report and corrections verbatim, separate from investigation interpretations. Record observed and expected behavior, location, trigger, conditions, and onset when known. Resolve target-changing uncertainty from available evidence; ask the user only for a fact the investigation cannot obtain, and continue unaffected work.