agentleFS
vindreshsinghvindreshsingh/engineering-skills/skills/orchestrated-delivery/SKILL.md

orchestrated-delivery

Skill0 starsChanged 5 months agoMITView on GitHub

Drives a feature end-to-end through the full lifecycle — define, plan, build, verify, review, ship — loading the right skill and persona at each phase, gating on sign-off, and dispatching independent work to parallel subagents. Use for any non-trivial feature or project, not a single small edit.

3 other projects use a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

7 rules. 0 are common, 6 are distinctive, and 1 is very specific to this project.

  • 0

    Common

    Most of the 3 edited versions we read keep these.

  • 6

    Distinctive

    This author's own choices. Worth a look.

  • 1

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    When you want the persona org (agents/sdlc/…) to act as a coordinated team, not a roster

    line 21
  2. Distinctive:

    Run the phases in order. Each phase has a governing skill, a persona lens, and an exit gate.

    line 28
  3. Distinctive:

    Do not advance until the gate is met.

    line 29
  4. Distinctive:

    Gate: every task is implemented and individually verified before it's marked done.

    line 51
  5. Distinctive:

    Gate: no blocking findings; implementation conforms to the spec and the agreed design.

    line 60
  6. Distinctive:

    No sign-off at the Define/Plan gates — request to shipped code in one unbroken run

    line 85
  7. Distinctive:

    No single source-of-truth plan.

    line 88

“Common” is measured: kept by at least 6 in 10 of the 3 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

Orchestrated Delivery

This is the conductor. Most failure on a real feature isn't bad code — it's jumping straight to code, skipping the spec, building the wrong thing, and discovering it at the end. Orchestrated delivery runs the work through the lifecycle as a system: at each phase it loads the governing skill, adopts the right persona, gates on exit criteria (and on your sign-off where it matters), and dispatches independent work in parallel — instead of one agent improvising the whole thing in a single pass.

This skill composes the others; it does not replace them. It decides which skill and which persona runs when, and refuses to advance a phase whose exit criteria aren't met.

When to Use

  • Any feature or project beyond a single small change
  • Work that spans disciplines (frontend + backend + data + ops) or multiple tasks
  • When you want the persona org (agents/sdlc/…) to act as a coordinated team, not a roster
  • Coordinating multi-step work that benefits from parallel execution

Skip for a one-line fix or a single isolated task — load that task's skill directly.

Process

Run the phases in order. Each phase has a governing skill, a persona lens, and an exit gate. Do not advance until the gate is met. Keep a single written plan as the source of truth throughout.

1. Define — what & why

  • Persona: product-manager / business-analyst. Skills: [[idea-shaping]] → [[spec-first]] (and [[product-brief]] for stakeholder sign-off).
  • Gate: a short spec exists with testable requirements and acceptance criteria, scope and out-of-scope named. Get the user's sign-off before planning.

2. Plan — how & in what order

  • Persona: solution-architect/technical-architect for design, then team-lead for breakdown. Skills: [[interface-design]], [[data-modeling]] for contracts; [[work-planning]] for the task plan.
  • Produce an ordered, dependency-aware task breakdown across disciplines, and mark which tasks are independent (parallelizable).
  • Gate: the plan is written, dependency-ordered, each task has a "done" check. User signs off.

3. Build — implement the plan

  • Execute task by task with [[incremental-delivery]] + [[test-first]]. Adopt the matching developer persona per task (frontend-developer, backend-developer, database-engineer) and apply the domain skill it needs ([[hardening]], [[resilience]], [[react-patterns]], [[accessibility]], [[caching-strategy]]…).
  • Dispatch independent tasks to parallel subagents ([[parallel-subagents]]); keep dependent tasks sequential. Update the source-of-truth plan as each task completes.
  • Gate: every task is implemented and individually verified before it's marked done.

4. Verify — prove it works

  • Skills: [[browser-checks]] for UI, run the test suite, check the spec's acceptance criteria.
  • Gate: acceptance criteria pass against the running system, not by assertion.

5. Review — correctness & safety

  • Persona: code-reviewer, security-auditor, technical-qc. Skills: [[review-gate]], [[hardening]], [[perf-budget]] where relevant.
  • Gate: no blocking findings; implementation conforms to the spec and the agreed design.

6. Ship — release safely

  • Persona: release-manager. Skills: [[launch-readiness]] (rollout, monitoring, rollback), [[migration-path]] if a contract/schema changed.
  • Gate: gradual rollout planned, rollback tested, monitoring live.

7. Operate (as needed)

  • [[observability]] in place; incident-commander + [[incident-response]] on standby.

Throughout: show work in digestible chunks and stop at the Define and Plan gates for sign-off — don't barrel from request to shipped code in one unbroken run.

Common Rationalizations

  • "Just start coding, we'll plan as we go." — Planning-as-you-go is how you build the wrong thing twice.
  • "The spec is obvious, skip to build." — Then the gate is free; write it and move on in two minutes.
  • "Do the whole thing in one pass." — A single undifferentiated run skips the gates that catch mistakes early.
  • "Run the independent tasks one at a time." — Wasted wall-clock; dispatch them in parallel.
  • "Verify at the very end." — Phase gates catch failures while they're cheap, not after they compound.

Red Flags

  • Build started with no agreed spec or plan
  • One giant task instead of a dependency-ordered breakdown
  • No sign-off at the Define/Plan gates — request to shipped code in one unbroken run
  • Independent tasks executed serially when they could parallelize
  • A phase advanced with its exit criteria unmet
  • No single source-of-truth plan; nobody can say what's done

Verification

  • Phases ran in order; each phase's exit gate was met before advancing
  • Spec and plan were signed off before Build
  • The governing skill and persona were loaded at each phase/task
  • Independent tasks were dispatched in parallel; dependent ones sequenced
  • Each task was verified before being marked done; acceptance criteria checked on the running system
  • Reviewed across the five lenses and shipped via [[launch-readiness]] with a tested rollback
  • A single written plan was kept current throughout

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.