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
- pa001024/dna-buildertheir own version, first seen Nov 9, 2026
- kynn-dev/agent-skillstheir own version, first seen Sep 30, 2026
- BlakeHastings/claude-code-artifactstheir own version, first seen Jun 13, 2026
Other versions3 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 3
- pa001024/dna-builder★ 155same name
- BlakeHastings/claude-code-artifacts★ 1same name
- kynn-dev/agent-skills★ 1same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository68 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- .github/copilot-instructions.mdCopilot instructions
- .cursor/rules/engineering-skills.mdcCursor rule
- marketing-routerSkill
- accessibilitySkill
- agent-guardrailsSkill
- agent-memorySkill
- agent-verificationSkill
- autonomous-loopsSkill
- browser-checksSkill
- caching-strategySkill
- context-curationSkill
- contract-testingSkill
- data-modelingSkill
- decision-docsSkill
- dependency-hygieneSkill
- design-handoffSkill
- e2e-testingSkill
- experimentationSkill
- fault-recoverySkill
- feature-flagsSkill
- finops-budgetSkill
- git-flowSkill
- hardeningSkill
- i18n-l10nSkill
- idea-shapingSkill
- incident-responseSkill
- incremental-deliverySkill
- interface-designSkill
- launch-campaignSkill
- launch-readinessSkill
- llm-feature-engineeringSkill
- long-running-agentsSkill
- community-engagementSkill
- content-marketingSkill
- email-nurtureSkill
- growth-strategySkill
- paid-adsSkill
- referral-loopSkill
- seo-growthSkill
- social-distributionSkill
- micro-interactionsSkill
- migration-pathSkill
- mobile-patternsSkill
- observabilitySkill
- parallel-subagentsSkill
- perf-budgetSkill
- pipeline-opsSkill
- product-briefSkill
- product-discoverySkill
- product-groomingSkill
- react-patternsSkill
- resilienceSkill
- review-gateSkill
- simplifySkill
- skill-creatorSkill
- skill-harvestSkill
- skill-routerSkill
- source-firstSkill
- The first 60 shown.
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
- Very specific: line 21
When you want the persona org (agents/sdlc/…) to act as a coordinated team, not a roster
- Distinctive: line 28
Run the phases in order. Each phase has a governing skill, a persona lens, and an exit gate.
- Distinctive: line 29
Do not advance until the gate is met.
- Distinctive: line 51
Gate: every task is implemented and individually verified before it's marked done.
- Distinctive: line 60
Gate: no blocking findings; implementation conforms to the spec and the agreed design.
- Distinctive: line 85
No sign-off at the Define/Plan gates — request to shipped code in one unbroken run
- Distinctive: line 88
No single source-of-truth plan.
“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.
Files
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-architectfor design, thenteam-leadfor 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
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
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.

