菜单
Produce or consume the canonical file-backed Mastermind executor report parsed by `run-task --post-only`. Use for executor evidence, schema-v1 report validation, controller audit inputs, or strict auditor advisory output.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
Turn a design handoff into a task contract that can actually be checked — named source, resolved components, token names instead of raw values, and an explicit list of what stays unverifiable. Use when a Figma frame, mockup, or design spec is handed over for implementation.
Convert a PRD, ticket, or feature request into a task contract that can fail — behaviour split from outcome, product nouns resolved to symbols that exist, unspecified cases surfaced as questions, and success metrics parked where a merge gate cannot wave them through. Use when product writing is handed over for implementation.
Enumerate what a change can reach and what enforces it — privileged operations and their callers, guards and the sites that apply them, and every module that reads a secret — before designing an auth, permission, or secrets change. Feeds the security review; produces a population and a gap list, never a verdict.
Read-only review of whether the tests in a finished change actually prove its behaviour — changed code no test reaches, a test exercising a different path than the one that changed, an assertion edited to match the new output, and a suite that ran nothing. Use after implementation, and whenever a change claims to be covered.
Gather who already depends on a service, handler, or state owner before changing it, and name the runtime gaps the codegraph cannot span. Use before designing a change to an API, queue consumer, background job, migration, or shared state; it feeds the architecture review rather than replacing it.
Read-only review of a finished UI change against the codegraph — a component nothing renders, a props contract changed without its callers, a reinvented component, or a raw value where the design system has a token. Use after implementing React or Vue work; triggers "review this UI change", "check the component", or a finished frontend diff.
| name | mastermind-structured-report-contract |
| description | Produce or consume the canonical file-backed Mastermind executor report parsed by `run-task --post-only`. Use for executor evidence, schema-v1 report validation, controller audit inputs, or strict auditor advisory output. |
| metadata | {"version":"0.3.0","authors":["mastermind"],"tags":["workflow","reporting"]} |
The only canonical machine-consumed agent handoff is
<task>/executor-report.md. The Rust controller reads that file, extracts its
sentinel-wrapped YAML, validates schema v1, and uses its supported claims and
verification observations during post-flight.
A copy returned in chat is useful to humans but is not the lifecycle record. The controller does not scrape planner chat, automatically patch a spec, or re-spawn an executor. Missing or malformed file evidence fails post-flight.
Append this tail to the prose report in executor-report.md:
<!-- mastermind:report-begin -->
```yaml
schema_version: 1
spec: .mastermind/tasks/<NNN>-<name>/spec.md
status: complete | partial | failed
phases:
- id: plan-1
status: done # done | pending | stopped_here | skipped
files_modified:
- path/relative/to/repository
claims:
- kind: function_added
symbol: <new symbol>
file: path/to/file
signature: "<indexed signature when available>"
- kind: integration
from: <changed caller>
from_file: path/to/caller
to: <existing callee>
to_file: path/to/callee
relation: calls
defects:
- kind: <recommended label or unclassified>
phase: plan-2
details: <concrete failure evidence>
remediation_hint: <bounded next action>
verifications:
- cmd: "<command actually run>"
result: pass | fail
output_excerpt: "<short excerpt on failure>"
observed:
exit_code: 0
tests_run: 12
```
<!-- mastermind:report-end -->
phases is the schema-v1 compatibility name for execution steps. New specs may
use IDs such as plan-1; they do not need phase-shaped prose or checklists.
The machine source of truth is
schemas/executor-report-v1.schema.json plus the stricter Rust consistency
checks. Unknown fields, unsupported versions, duplicate step IDs, empty
commands, contradictory complete reports, and reports over 1 MiB fail closed.
complete: every reported step is done, defects is empty, and every
verification result is pass.partial or failed: at least one concrete defect is required.files_modified: the executor's evidence, not the scope authority. The
controller calculates the real changed-file set from git and compares it with
the spec.claims: only function_added and integration are understood by the
deterministic audit. Use claims: [] for general prose outcomes.defects[].kind: a recommended routing label from defect-taxonomy.md, not a
closed enum and not an instruction for automatic repair.verifications: record commands actually run. Do not claim an inferred or
intended check.Strict mode may ask a read-only auditor for a sentinel-wrapped advisory tail:
<!-- mastermind:audit-begin -->
```yaml
spec: .mastermind/tasks/<NNN>-<name>/spec.md
verdict: held | drift | broken
scope_match: true
discrepancies: []
verifications_rerun: []
```
<!-- mastermind:audit-end -->
This tail is currently a human/planner review contract, not a Rust-parsed
lifecycle input. audit.md and state.json remain controller-owned. Do not
claim that an auditor chat message was persisted unless a future controller
command actually implements that transition.
The planner reads malformed, partial, or failed evidence and decides whether to repair implementation, revise the contract, or stop. Recommended defect labels help grouping; they do not authorize deterministic auto-fixes. After three cycles blocked by the same condition, return to design or the user.