agentleFS
JipletJiplet/transformation-os/skills/pipeline/clevel-review/SKILL.md

clevel-review

Skill2 starsChanged 5 months agoMITView on GitHub

Pre-output gate for senior-audience deliverables. Auto-fire ONLY when producing a decision brief, one-pager, conversation prep, exec summary, or "support asks" document bound for a named GM, CFO, Sector Exec, ExCo, or external client (government department, telco client, Defence, national network client, etc.). Forces routing through the c-level thinker council before any write-up. Refuses to produce final output without classifying audience, loading thinker files, and running council critique. SKIP for routine wiki edits, internal team docs, status updates, or any deliverable not bound to a named senior audience.

The file

We split this skill into the rules it gives an agent

8 rules. About 0 are common, 5 are distinctive, and 3 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈5

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 3

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Ask the user (via AskUserQuestion if any are missing):

    line 16
  2. Very specific:

    Read <workspace>/CLAUDE.md "Orchestrator, Routing Logic" section (lines covering thinker triggers).

    line 26
  3. Very specific:

    Read each matched thinker file from <workspace>/context/<thinker>.md.

    line 28
  4. Distinctive:

    Stop: Ask user to confirm before Phase 2: "Council looks right?

    line 64
  5. Distinctive:

    Stop: Ask user to confirm or push back on any critique before Phase 3.

    line 102
  6. Very specific:

    Auto-chain: after Phase 3 produces the draft, automatically invoke /exec-comms to apply visual + structural format discipline (callouts, …

    line 119
  7. Distinctive:

    Audience unknown: ASK via AskUserQuestion. Do not guess.

    line 164
  8. Distinctive:

    Multiple audiences: default to most senior; flag secondary as constraint.

    line 165

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

/clevel-review

Pre-output gate. Three phases. Hard enforcement: do NOT skip to Phase 3.

Canonical spec: <workspace>/6 - slash commands/clevel-review.md

Phase 1: Classify and self-validate the council

Required inputs

Ask the user (via AskUserQuestion if any are missing):

  • Deliverable type — decision brief / status / risk / category strategy / contract change / conversation prep / one-pager / steerco pack / external client doc
  • Audience — named person + role (e.g. "Group GM Supply Chain"; "CFO"; "client contact at a government department")
  • Format — 1-page Obsidian markdown / Word docx / PPTX deck / dashboard / email
  • Ask type — support / approval / information / decision / endorsement
  • Source material — brief, draft, wiki path, prior outputs (if any)

Routing

  1. Read <workspace>/CLAUDE.md "Orchestrator, Routing Logic" section (lines covering thinker triggers).
  2. Match the inputs above to thinker triggers. Multi-signal tasks load multiple thinkers.
  3. Read each matched thinker file from <workspace>/context/<thinker>.md.

Mandatory thinker triggers (not exhaustive, see CLAUDE.md routing table for full):

SignalThinker file(s)
Initiative sizing / "should we do this"cso.md
Financial claim / one-pager / CFO summarycfo.md
Named client (a state-government facilities client, a telco client, Defence, a national network client, state transport)chief-client-officer.md + general-counsel.md
Category, vendor, panel, supplier consolidationcpo.md + coo.md
System / data / Databricks / SAP dependencycio.md + data-landscape.md
Contract terms, EA, legal feasibilitygeneral-counsel.md + cro.md
Workforce, leave, restructurechief-people-officer.md + general-counsel.md
Risk, insurance, WHS, regulatorycro.md
Operational feasibility, change managementcoo.md + delivery-difficulties.md
Cost modelling, inflation, market rateschief-economist.md
Visual deliverable (deck, dashboard, web)cdo.md
Steerco / ExCo updatecso.md + cfo.md

Phase 1 output (write to user)

## Classification
| Field | Value |
|---|---|
| Deliverable type | … |
| Audience | … |
| Format | … |
| Ask type | … |

## Council loaded
| Thinker | Why loaded | File read |
|---|---|---|
| … | … | ✓ |

Stop. Ask user to confirm before Phase 2: "Council looks right? Add or remove any thinker before I run critique."

Phase 2: Council critique

For each loaded thinker, produce critique IN THAT THINKER'S VOICE using:

  • The thinker's "Questions This Thinker Always Asks" (apply to the draft / proposal)
  • The thinker's "Anti-Patterns This Thinker Catches" (red-flag scan)
  • The thinker's "Company Context" (relevance check)

Output format

## Council critique

### [Role] (file: [thinker].md)

**Voice:** [1-2 sentences framing what this thinker brings]

**Challenges raised:**
- [Specific challenge tied to the draft / proposal]
- …

**Anti-patterns flagged:**
- [Specific pattern in the draft this thinker catches]

**Yes if:**
- [What would need to change for this thinker to support]

---

[Repeat per thinker]

## Consolidated MECE issues
| # | Issue | Raised by | Severity | Resolution |
|---|---|---|---|---|
| 1 | … | … | high/med/low | … |

Stop. Ask user to confirm or push back on any critique before Phase 3.

Phase 3: Audience-shaped write-up

Format selected from classification, NOT from topic.

Format library (v1)

AudienceFormatSkill to call
GM / Sector exec1-page Obsidian markdown(this skill, GM template below)
CFO1-page markdown + optional Wordprogram-exec-summary
ExCo / SteercoPPTXprogram-steerco-pack
External clientWord docanalysis-commercial-insight

v1 ships GM-only. For CFO / ExCo / External, run Phase 1 + 2 then hand off to the named skill with the consolidated issue list.

Auto-chain: after Phase 3 produces the draft, automatically invoke /exec-comms to apply visual + structural format discipline (callouts, parallel tables, bold-the-punchline, banned-word scan, length cap to audience tier). The chain is mandatory, not optional. /exec-comms enforces the format rules; /clevel-review produces the content.

GM 1-page template (mandatory structure)

# [Title]

> Audience: [named person] | Read time: [N] min | Date: [YYYY-MM-DD]

## Decision sought
[1 sentence. The single thing this page asks for.]

## Situation
- [3-4 MECE bullets. Where we are.]

## Complication
- [3-4 MECE bullets. Why current state cannot deliver alone.]

## Asks (parallel)
| # | Ask | Who carries if approved | Effort | Outcome |
|---|---|---|---|---|

## What we commit in return
- [3 bullets. Reciprocal commitment.]

## What's at stake by [date]
- [2 bullets. Cost of inaction, anchored to a real deadline.]

## Appendix
[Links only. No narrative.]

Banned patterns (refuse to ship if any present)

  • "What this is NOT" sections (defensive, undermines confidence)
  • "Pre-empted Questions" sections (if asked, answer; don't pre-litigate)
  • Mixed tactical / structural detail in same section (force separate buckets)
  • Asks of different shapes (parallel structure mandatory)
  • Value tables before the ask (lead with the answer, not the proof)
  • Headlines without urgency markers (anchor to deadline)

Safety / failure modes

  • Thinker file missing: warn, list missing thinker, proceed with degraded council, flag in output.
  • User wants to skip Phase 2: warn that this defeats the gate; require explicit --skip-critique flag from user; log the skip in the output deliverable.
  • Audience unknown: ASK via AskUserQuestion. Do not guess.
  • Multiple audiences: default to most senior; flag secondary as constraint.
  • Conflict between thinkers: surface to user as decision point. Do not auto-resolve.

Logging

Optional. If user passes --log <path>, write the Phase 1 + 2 outputs as a markdown file at that path alongside the deliverable. Default: no log.

Reference

  • Spec: <workspace>/6 - slash commands/clevel-review.md
  • Routing logic: <workspace>/CLAUDE.md (Orchestrator section)
  • Thinker files: <workspace>/context/*.md
  • Output skills: program-exec-summary, program-steerco-pack, analysis-commercial-insight

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.