machenjie engineering-change-analysis
Role
Support analysis-agent in the selected source-backed mode. For Analyzed Work,
the current Engineering Brief is the only operational analysis authority;
Specialist work is input pending Brief incorporation.
When To Use
- implementation-preparation for source-backed desired-behavior, affected-consumer, and verification analysis before implementation or repair
- diagnosis-only for verified cause analysis
- source-backed-answer for a repository-dependent engineering question
Do Not Use
- eligible bounded Direct Task
- source-free question
- accepted Engineering Brief already exists and the user explicitly requests one narrow artifact analysis
- module-boundary placement decision
- dependency-direction placement decision
Required Inputs
- selected mode
- bounded source evidence and constraints
Professional Decision Rules
- Bind the selected mode to its
mode-contract. - Apply Core
task_contract.analyzed_work_authorityfor initial closure, decision-invalidated Delta Analysis, transitive impact, and Skill routing. - Prove its owner, impact, failure, validation, and proof limits.
- Put source-proven placement and Specialist results in the Brief.
- Route unresolved structural placement to
architecture-impact-reviewer. - Put a complete Task Contract v2 that Main can dispatch verbatim in the Brief, not a derived Task DAG or handoff.
High-Value Gotchas
- Proximity is not ownership.
Execution Checklist
- Load the selected mode contract and triggered guidance.
- Return its result and proof limit under Core.
Stop / Escalation Conditions
- Stop for user decisions or result-invalidating gaps.
- Brief conflict or protected change returns
blocked. - Structural placement or dependency direction routes to
architecture-impact-reviewer. - Foundational invalidation permits full analysis.
Output Contract
- mode result, source evidence, and proof limit
- implementation-preparation: authoritative Brief, dispatchable Slice, Specialist input
Targeted References
| Path | Type | Load when | Do not load when | Required by | Required output |
|---|---|---|---|---|---|
| checklist | decision-checklist | A bounded review needs a concrete surface inventory across product, UX, domain, API, data, frontend, backend, integrations, security, tests, release, observability, and docs | Evidence freshness, consumer proof, or performance blast radius is the core risk | analysis-agent | checklist-result, residual-risk |
| diagnosis only | mode-contract | engineering change analysis selects diagnosis-only | engineering change analysis is executing another mode | analysis-agent | mode-result, proof-limit |
| evidence patterns | evidence-pattern | Closure depends on changed-surface-to-validation mapping, same-pattern scan proof, consumer evidence, rollback limits, graph/report freshness, or proof limits | Only high-level routing is needed and evidence is not being closed | analysis-agent | evidence-record, proof-limit, residual-risk |
| implementation preparation | mode-contract | engineering change analysis selects implementation-preparation | engineering change analysis is executing another mode | analysis-agent | mode-result, proof-limit |
| index | index | competing engineering change analysis references require dependency, conflict, or output-fragment selection | the engineering change analysis root or a task-named reference already resolves selection | analysis-agent | reference-selection |
| solution optimality | targeted | Performance/resource blast radius may include CPU, memory, network, disk, lock contention, throughput, or latency | No performance-sensitive path or resource tradeoff is material | analysis-agent | selected-approach, residual-risk |
| source backed answer | mode-contract | engineering change analysis selects source-backed-answer | engineering change analysis is executing another mode | analysis-agent | mode-result, proof-limit |