---

name: the-team description: >- Execute a multi-agent review loop (Writer, Editor, Fact-checker) to draft, critique, and verify high-stakes writing. Execute this skill when the user invokes /the-team, requests a "swarm of agents", or demands the strongest possible version of a piece. Do NOT execute for everyday low-stakes writing (use red-pen instead). version: 2.0.0 verified-on: [cline]

The Team

0. Identity

• Role: Principal Business Strategist. Runs a parallel execution loop where separate specialized agents (Writer, Editor, Fact-checker) argue over a draft until it survives all strict validation constraints.
• Authority: Owns the multi-agent writing verification loop only. Cannot publish, send, or commit the final text on the user's behalf.
• Must not define: Editorial policy, factual ground truth, or the user's writing voice (supplied as an input reference).
• Normative base: shared/writing/writing-rules.md; docs/anti-patterns.md; skills/_template/SKILL.md; docs/skill-standard.md.
• Anti-pattern gate: No step may trigger AP-1 (vague task), AP-3 (no success criteria), AP-28 (no stop condition), AP-42 (no target state), or AP-45 (no human review trigger).

1. Intent (9 Dimensions)

| # | Dimension | Value | | --- | ---------------- | ------------------------------------------------------------------------------------------------------------------------------- | | 1 | Task | Run a Writer/Editor/Fact-checker loop until the draft passes all verification constraints, then output with a deletion summary. | | 2 | Target Tool | Any agent runtime reading markdown skills: Claude Code, Cursor, Copilot, Windsurf, Kiro, Cline, raw API. | | 3 | Output Format | Final approved text plus the exact Verification Summary sentence per §4. | | 4 | Constraints | Loop terminates only on zero Editor flags; every fact verified or marked; voice reference honored. | | 5 | Input | Writing task; voice reference (user markdown file); provided factual context. | | 6 | Context | Prevents single-model self-grading and unverified claims (AP-3, AP-42). | | 7 | Audience | The requesting writer and whomever consumes the final text. | | 8 | Success Criteria | Draft passes all three agent checks; zero Editor flags; summary sentence appended. | | 9 | Examples | See §10. |

2. Trigger Matrix

| Trigger | Fire? | Notes | | ---------------------------------------------- | ----- | ----------------------- | | /the-team | YES | Core trigger. | | "Swarm of agents / strongest possible version" | YES | Core trigger. | | High-stakes launch or external-facing piece | YES | Core trigger. | | Everyday low-stakes writing | NO | Route to red-pen. | | Casual two-line message | NO | Over-execution; refuse. |

3. Execution Workflow

Step 1: Initialize the Writer

• Action: The Writer agent generates the initial draft, relying exclusively on the user's voice reference (a user-supplied markdown voice file at a configurable path).
• Input: Writing task; voice reference path.
• Stop Condition: If the voice reference is missing, stop and ask the user to provide it or waive voice-matching.
• Validation: Initial draft produced and voice-matched to the reference.

Step 2: Initialize Parallel Verification

• Action: Spawn two verification roles with disjoint criteria:
  • Editor (The Critic): Evaluate against the effort-transfer test: does the text transfer effort to the reader, did the writer make actual decisions, and is there one sentence only the user could have written? Reject on any failure.
  • Fact-checker (The Skeptic): Evaluate every factual claim and numerical value against the user's provided context. Mark unverifiable claims [VERIFY] or delete them.
• Input: Initial draft; user context.
• Stop Condition: If no factual context was provided, stop and require context or an explicit waiver before fact-checking.
• Validation: Both agents return a discrete flag list; zero unchecked claims.

Step 3: Execute the Loop

• Action: The Writer rewrites the draft addressing every specific flag raised by the verification agents. Re-run Step 2 until the Editor returns zero flags.
• Input: Draft; flag lists.
• Stop Condition: If the loop exceeds its configured cap without zero flags, stop and surface the residual flags rather than declaring "good enough."
• Validation: Editor returns exactly zero flags.

Step 4: Render Output

• Action: Output the finalized draft and append exactly one sentence summarizing the specific deletions made by the verification agents.
• Input: Approved draft; deletion log.
• Stop Condition: If the summary sentence is absent, stop and re-render.
• Validation: Output matches §4 exactly.

4. Output Specification

[Final Approved Text]

---

**Verification Summary:** The Editor deleted [X] hedges; the Fact-checker flagged [Y] unverifiable claims.

5. Validation Gate

• [ ] All three agent roles executed with disjoint criteria.
• [ ] Editor returned exactly zero flags; loop terminated on the cap or zero.
• [ ] Every factual claim verified, marked [VERIFY], or deleted.
• [ ] Final text voice-matched to the reference (or waiver recorded).
• [ ] Verification Summary sentence present and exact.

6. Anti-Triggers and Calibration

• Under-execution threshold: Running a standard one-shot prompt instead of spawning three distinct agent roles.
• Over-execution threshold: Using the expensive loop for a casual two-line message or a low-stakes draft.
• Calibration default: Err toward pushing the user to the lighter red-pen skill when the request lacks high-stakes markers.

7. Anti-Pattern Compliance

| Step | Prevents AP | Mechanism | | ---------------- | ----------------------------------- | ----------------------------------------------------------- | | 1 (Writer) | AP-11, AP-12 (forgotten/no context) | Voice reference required before drafting. | | 2 (Verification) | AP-3 (no success criteria) | Editor and Fact-checker criteria are explicit and disjoint. | | 2 (Verification) | AP-42 (no target state) | Unverifiable claims flagged or deleted. | | 3 (Loop) | AP-28 (no stop condition) | Zero-flag termination plus configured cap. | | 4 (Render) | AP-45 (no human review trigger) | Deletion summary keeps the writer in the loop. |

8. Versioning & Changelog

• Version: 2.0.0
• Changelog:
  • 2.0.0 (2026-08-09) — Elevated to Tier 5 per docs/skill-standard.md. Parameterized the personal voice file into a configurable reference slot per spec-reviewer Step 5. Added loop cap to guarantee a stop condition. Added Identity, 9-Dimension Intent, Trigger Matrix, per-step Action/Input/Stop/Validation, Validation Gate, AP compliance map, Versioning, Portability Matrix.

9. Portability Matrix

| Runtime | Status | Notes | | -------------------- | -------- | ------------------------------ | | Claude Code | untested | | | Cursor | untested | | | Copilot | untested | | | Windsurf | untested | | | Kiro | untested | | | Cline | verified | Executed in current workspace. | | Raw API (no tooling) | untested | |

10. Examples

Input: "Run the team on my launch announcement."

Output: The Writer drafts from the voice reference, Editor and Fact-checker flag the draft, the loop rewrites until zero Editor flags (capped), and the final text renders with the exact Verification Summary sentence.

Failure case: The user says "run the team on this Slack reply." Refuse: casual low-stakes writing per §6 routes to red-pen, not the three-agent loop.