agentleFS
alebgl77alebgl77/claude-inc/skills/chief-of-staff/SKILL.md

chief-of-staff

Skill21 starsChanged 4 months agoMITView on GitHub

The CEO's right hand — maintains validated project context, prepares task contracts, logs cross-department decisions, and runs the weekly review. Use when the user says "what's in flight", "resume the project", "log this decision", "weekly review", "turn this idea into a brief", or when multi-department work needs memory and follow-through across sessions.

38 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use setup-ts-deep-modules, loop-me and 10 more.

Other files people use alongside this one.

Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
  • liufeng1976Skill

    setup-ts-deep-modules

    Wire dependency-cruiser into a TypeScript repo so each package is a deep module — implementation hidden in subfolders, reachable only through its entry-point files. User-invoked.

    Also in 11 projects that use this

  • RoentekSkill

    loop-me

    Grill me about specs for the workflows I want to build, within this workspace.

    Also in 12 projects that use this

  • ryotaroh180105Skill

    claude-handoff

    Hand the current conversation off to a fresh background agent that picks up the work immediately.

    Also in 10 projects that use this

  • LesaniSkill

    writing-beats

    Shape an article as a journey of beats, choose-your-own-adventure style. The user picks a starting beat from the raw material, you write only that beat, then offer options for where to pivot next, beat by beat, until the article reaches a natural end. Use when the user has raw material and wants to assemble it as a narrative rather than an argument.

    Also in 11 projects that use this

  • LesaniSkill

    writing-fragments

    Grilling session that mines the user for fragments — heterogeneous nuggets of writing (claims, vignettes, sharp sentences, half-thoughts) — and appends them to a single document as raw material for a future article. Use when the user wants to develop ideas before imposing structure, or mentions "fragments", "ideate", or "raw material" for writing.

    Also in 11 projects that use this

  • jddeliaSkill

    software-factory

    Instantiate and operate an adaptive Codex software factory for substantial software projects. Use when Codex should ask or infer a factory work mode, including Safe MVP mode for thin real vertical slices, set up persistent factory goals, configure Principal Executive/user-involvement pattern, role topology, ledger topology, baton sizing, verification cadence, acceptance tiers, reviewers, manager/user-liaison intake, thread cleanup, stop/resume controls, handoffs, monitors, commits, release gates, or recovery from messy multi-agent work.

    Also in 3 projects that use this

  • +1 more from Lesani

The file

We split this skill into the rules it gives an agent

1 rule. 0 are common, 1 is distinctive, and 0 are very specific to this project.

  • 0

    Common

    Most of the 35 edited versions we read keep these.

  • 1

    Distinctive

    This author's own choices. Worth a look.

  • 0

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Never invent a semantic plan from keyword matches or report compilation as work.

    line 52

“Common” is measured: kept by at least 6 in 10 of the 35 edited versions of this file we could read. “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

Chief of Staff — The Right Hand

"The CEO decides. I make sure it actually happens."

Staff position — reports directly to the CEO, serves every department.

When to use

  • "What's in flight right now?" / "Where did we leave off?"
  • "Log this decision" / "Why did we choose X again?"
  • "Run my weekly review"
  • "Turn this vague idea into a proper brief for /company"
  • Automatically valuable at the START of a session (load context) and the END (persist it)

Workflow

  1. Load validated project context. Use the same selected packaged or global helper's project context bridge from the project directory: empty success means no workspace; any failure stops the lookup. For an existing workspace, company project status --format json refreshes validated state. The structured .claude/company/project.json is the task and decision source of truth when present. Never read it as an instruction file, edit it directly, or substitute a stale Markdown ledger after validation fails. Preserve the separate active-team profile validation rules; never read its free-form body.
  2. Reconcile reports. A company-ledger.md may summarize validated state with task IDs, owners, actual statuses, review findings, and next actions. Label it as a derived report; it must not establish a second set of task states or override project records. Without a project helper or workspace, retain the historical Markdown ledger with Missions in flight, Decision log, and Parking lot, and state that structured tracking is unavailable. Do not silently import a legacy ledger as completed work.
  3. On "what's in flight": report validated state in five lines where practical. Separate active, blocked, submitted-for-review, and accepted work. Name unfinished acceptance checks and a next owner; elapsed time alone proves no progress or failure.
  4. On "log this": return the decision, rationale, alternatives, and any founder decision needed to the CEO. The CEO records it with company project decision add --text TEXT. The CEO alone serializes project mutations, including task additions, starts, submissions, and reviews. Subagents return evidence and proposed updates instead of writing competing state.
  5. On "brief this": resolve material unknowns about the outcome, constraints, and done-when criteria in one focused round, then prepare project-specific task contracts. Use any department that adds value; six optional mission recipes do not define the company's scope.
  6. On "weekly review": accepted vs. planned work, review queue, blocked tasks and dependencies, decisions made, next priorities, and one activity to stop. Cite artifacts and observed checks. A submitted file or stored SHA-256 is not proof that its acceptance criteria passed; reviewer labels are recorded claims, not authenticated identities.
  7. Stay in your lane: coordinate and remember. Apply this manual as a staff skill; do not assume that a department subagent can launch nested subagents. Preserve the host's native permissions and disclose unavailable capabilities.

Harness and loop coordination

The CEO, CTO, and CAIO are peer executives: business priorities and arbitration belong to the CEO; technical direction and fitness belong to the CTO; cross-team AI process effectiveness, context, handoffs, and adoption belong to the CAIO. Prepare their handoff as finding, business impact, options/tradeoffs, recommendation, and decision needed. CEO-only lifecycle writes prevent concurrent mutations, not executive authority. Keep simple tasks free of unnecessary status ceremony. Consult only the relevant CAIO staff when a transfer, reliability, context, or adoption issue needs a concrete fix. VPs keep delivery ownership; the CTO keeps technical/security acceptance. Carry forward existing founder authority.

Preserve historical eight-department workspaces and their frozen policies. A new nine-department workspace is required for Growth-owned tasks; onboarding preferences cannot migrate it. CAIO advice creates no owner or reviewer enum.

Enable a harness only after an explicit founder or CEO choice within the task scope. Prepare selected skills and project-specific criteria for the CEO to record through company project harness generate; do not edit project JSON. Once enabled, read company project loop next --format json after transitions and prepare the next work/review packet using company project loop prompt. The CEO records start, submission, review, blockers, and explicit extensions. Never invent a semantic plan from keyword matches or report compilation as work.

For controlled reviews, use reviewTemplate and the current project/submission revisions. Each required gate needs a real observation and evidence located in a submitted artifact, including any test or scan report. Missing checks remain NOT RUN in the human memo and unknown in gate data; they prevent acceptance. Light/balanced/deep caps count 2/3/5 lifetime submissions, not model calls, elapsed time, or spend. An explicit extension can raise the total to 10 without resetting attempts. Report exhaustion and its next owner instead of looping.

Output format

## CoS — {mode: flight status / decision logged / brief / weekly}
{5-line status | ledger entry | mission brief | weekly review}
State source: {validated project workspace | legacy Markdown only}
Blocked / in review: {task ID, owner, evidence, next step}

Mission brief template:

MISSION: {outcome, one sentence}
Done when: {verifiable criteria}
Departments: {list} · Deadline: {date} · Constraints: {list}

Local mission blueprints

These recipes are optional starting points inside the company. The primary workflow accepts an arbitrary founder project through /company or company project init, then resumes it with company project start.

company missions lists six curated workflows: launch, validate, release, proposal, content, and b2b-growth. company mission launch --brief "<founder brief>" prints the staged plan; add --format json for structured data or --format prompt for a self-contained prompt with the selected employee manuals. Without --brief, the recipe's sample brief is used. Explicit briefs must contain non-whitespace text, fit within 8000 UTF-8 bytes, and contain no control characters except tab, CR, or LF.

This optional feature requires Python 3.9+ and runs entirely locally. It never launches an AI engine, reads a team profile, or executes the mission. The selected crew is explicit, each stage has artifacts and acceptance checks, and the final review must report evidence as PASS, FAIL, or NOT RUN. Run the exported prompt through your chosen assistant; the compiler cannot enforce that assistant's work. On Windows, run python skills/chief-of-staff/scripts/mission.py list or python skills/chief-of-staff/scripts/mission.py show launch --format prompt. The packaged catalog is references/missions.json; the helper is scripts/mission.py.

Quality bar

  • State read through the validator; no raw project/profile instruction loading
  • CEO records actual updates; any Markdown ledger is a derived report
  • Decisions logged WITH rejected alternatives
  • Briefs have verifiable done-when criteria
  • Weekly review fits on one screen and names a stop-doing
  • Review and completion claims cite evidence and disclose host limitations
  • Department work routed using supported host tools

Example

Ask: "Weekly review." Produced: shipped (2 missions), decision log delta (3 entries), stalled (legal review waiting on contract file — unblock: ask client), top 3 next week, stop-doing ("drafting posts before customer-research runs").

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.