Delegation Protocol
The chain
Architect decides. Engineer implements. Quality verifies. Release ships.
Never skip a link. Implementation with no design decision produces rework.
A release with no verification produces an incident. Both cost more than the link you skipped.
Routing procedure
- Name the domain in one sentence. If you need two sentences, it is two tasks — split first.
- Find the owner in
.claude/teams/README.md.
- Pick the minimum set. One agent per distinct concern. Not one per file, not one per idea.
- Gather the declared Inputs before delegating. An agent without its inputs guesses, and a
confident guess is worse than an admitted gap.
- Delegate with context, not with the user's raw prompt. State what was already decided,
what is out of scope, and what you want back.
- Reconcile. Where two agents disagree, do not average them. State the disagreement
explicitly and escalate to the team's architect with both positions.
Fan-out budget
| Task |
Agents |
Why |
| One file, obvious fix |
0 — do it inline |
Delegation costs more than the fix |
| One feature, one domain |
1-2 |
Implementer, then reviewer |
| Cross-domain feature |
3-4, one per domain |
Any more and they overlap |
| System design |
Architects first, implementers in a second pass |
Design before build, always |
More than four agents on one task means the task is not decomposed yet. Decompose it.
Worked example
"Our checkout times out under Black Friday load and we lose orders."
Wrong: spawn eight agents — backend, database, SRE, performance, QA, security, PM, release —
and merge eight overlapping reports.
Right — sequential, each gated on the previous:
1. performance-test-engineer → where does it actually break?
Returns: "p99 8.2s at 400 rps; DB connection pool saturated at 200 rps, not CPU."
→ Everything below is now scoped by that finding. Without it, the rest is guesswork.
2. database-architect → pool sizing, connection strategy, read replica options
backend-architect → idempotency so retries do not double-charge
(parallel — different concerns, same evidence)
3. backend-developer → implements the agreed design
database-engineer → pool config and index changes
4. code-reviewer → correctness of the retry path
reliability-test-engineer → verify it degrades instead of collapsing
5. release-manager → canary with abort criteria tied to p99
Five stages, nine agents total, but never more than two at once, and each stage
gets real inputs from the last. Step 1 alone eliminated four wrong directions.
Anti-patterns
| Anti-pattern |
What goes wrong |
| Spawning five agents to answer one question |
Five overlapping reports, no decision, high cost |
| Delegating without declared inputs |
Agent invents plausible numbers; you cannot tell |
| Skipping the architect on an irreversible decision |
Rework at 10x the cost, three weeks later |
| Accepting output without checking evidence |
You inherited a hallucination and now own it |
| Forwarding the user's raw prompt to the agent |
Agent re-derives context you already have |
| Parallel agents on the same concern |
They contradict; you arbitrate with no basis |
Done when
1---2name: delegation-protocol3description: How to route work across teams and agents in this repository. Use when a task spans more than one domain, when you are unsure which agent owns it, or before fanning out to several agents.4---56# Delegation Protocol78## The chain910**Architect decides. Engineer implements. Quality verifies. Release ships.**1112Never skip a link. Implementation with no design decision produces rework.13A release with no verification produces an incident. Both cost more than the link you skipped.1415## Routing procedure16171. **Name the domain** in one sentence. If you need two sentences, it is two tasks — split first.182. **Find the owner** in `.claude/teams/README.md`.193. **Pick the minimum set.** One agent per distinct concern. Not one per file, not one per idea.204. **Gather the declared Inputs** before delegating. An agent without its inputs guesses, and a21 confident guess is worse than an admitted gap.225. **Delegate with context**, not with the user's raw prompt. State what was already decided,23 what is out of scope, and what you want back.246. **Reconcile.** Where two agents disagree, do not average them. State the disagreement25 explicitly and escalate to the team's architect with both positions.2627## Fan-out budget2829| Task | Agents | Why |30|---|---|---|31| One file, obvious fix | **0** — do it inline | Delegation costs more than the fix |32| One feature, one domain | 1-2 | Implementer, then reviewer |33| Cross-domain feature | 3-4, one per domain | Any more and they overlap |34| System design | Architects first, implementers in a second pass | Design before build, always |3536More than four agents on one task means the task is not decomposed yet. Decompose it.3738## Worked example3940> "Our checkout times out under Black Friday load and we lose orders."4142**Wrong:** spawn eight agents — backend, database, SRE, performance, QA, security, PM, release —43and merge eight overlapping reports.4445**Right — sequential, each gated on the previous:**4647```481. performance-test-engineer → where does it actually break?49 Returns: "p99 8.2s at 400 rps; DB connection pool saturated at 200 rps, not CPU."5051 → Everything below is now scoped by that finding. Without it, the rest is guesswork.52532. database-architect → pool sizing, connection strategy, read replica options54 backend-architect → idempotency so retries do not double-charge55 (parallel — different concerns, same evidence)56573. backend-developer → implements the agreed design58 database-engineer → pool config and index changes59604. code-reviewer → correctness of the retry path61 reliability-test-engineer → verify it degrades instead of collapsing62635. release-manager → canary with abort criteria tied to p9964```6566Five stages, nine agents total, but **never more than two at once**, and each stage67gets real inputs from the last. Step 1 alone eliminated four wrong directions.6869## Anti-patterns7071| Anti-pattern | What goes wrong |72|---|---|73| Spawning five agents to answer one question | Five overlapping reports, no decision, high cost |74| Delegating without declared inputs | Agent invents plausible numbers; you cannot tell |75| Skipping the architect on an irreversible decision | Rework at 10x the cost, three weeks later |76| Accepting output without checking evidence | You inherited a hallucination and now own it |77| Forwarding the user's raw prompt to the agent | Agent re-derives context you already have |78| Parallel agents on the same concern | They contradict; you arbitrate with no basis |7980## Done when8182- [ ] The owning team is named and justified in one line.83- [ ] Each delegated agent received its declared Inputs.84- [ ] No two agents were given the same concern.85- [ ] Contradictions between agents are stated, not silently averaged.86- [ ] Every claim in the merged result traces to an agent's evidence, not to your summary of it.
bakhod1r