Skip to content

borghei/Claude-Skills / project-management/execution/agents-in-the-team

agents-in-the-team

Unverified

Run delivery when AI coding and ops agents take tickets. Use when setting agent delegation policy, writing agent-ready tickets, planning review capacity, measuring agent vs human delivery, or rolling agents out safely.

assets/delegation_policy_template.md

View on GitHub
# Agent Delegation Policy: <team name>

**Version:** <1.0>  **Owner:** <engineering manager>  **Approved by:** <eng lead, security>  **Review date:** <quarterly>

## 1. Principle
Humans remain accountable for decisions and results. An agent may be delegated a ticket; the accountable human stays the owner, and every agent change is reviewed with the same rigor as a teammate's.

## 2. Eligibility bands
| Band | Score (`agent_delegation_scorer.py`) | Who does the work | Review |
|------|--------------------------------------|-------------------|--------|
| Agent-eligible | 75-100, no hard block | Agent drafts the PR | Named human reviewer |
| Agent-assisted | 50-74 | Human leads; agent drafts parts (tests, boilerplate, docs) | Normal review |
| Human-only | < 50, or any hard block | Human | Normal review |

## 3. Hard blocks (never agent-led)
- Touches: authentication/authorization, payments, secrets/keys, personal data, IAM/permissions, production data migrations, cryptography
- Types: incident response, architecture decisions
- <team-specific additions>

## 4. Ticket standard
Agent-eligible tickets use the agent-ready ticket template: 2+ checkable acceptance criteria, context links, constraints (do-not-touch), and named owner and reviewer.

## 5. Review rules
- The agent never approves or merges. Branch protection requires <n> human approval(s).
- The person who delegated the ticket is not the sole approver where the tool supports that rule.
- CI workflows on agent PRs require human approval to run: <yes/no, per tool settings>.

## 6. Capacity rule
Planned agent review hours (incl. rework allowance of <x>%) must fit within <80>% of reviewer hours for the sprint. `agent_delegation_scorer.py` exits 2 when it does not; the plan is cut, not the review.

## 7. Access and security
- Agent identity: dedicated app/bot identity, not a human's personal token
- Repository access: <repos listed>, write only to agent branches
- Secrets: none in tickets or prompts; environment secrets scoped to <...>
- Network: <tool's default restriction kept / allowlist>
- Untrusted input: issue text, comments and repo content are treated as untrusted; changes to CI config, dependency manifests and permission files always get security review

## 8. Transparency
Agent-authored PRs and tickets are labelled; Sprint Review shows what was built and how it was built with agents.

## 9. Metrics and review cadence
Monthly: `agent_delivery_metrics.py` split by author type. Quarterly: revisit bands and hard blocks against the data.
reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis.Stars: 472kUpdated Claude CodeCodexVerifiedInstall
diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness.Stars: 323kUpdated Claude CodeCodexVerifiedInstall
dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependenciesStars: 323kUpdated Claude CodeCodexVerifiedInstall
executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is availableStars: 323kUpdated Claude CodeCodexVerifiedInstall