Subagent driven development
Skill yigityildiz0/universal-ai-skill-library/skills/archive-variants/subagent-driven-development
Execute an implementation plan through bounded implementer and independent reviewer agents with disjoint write ownership, explicit handoffs, evidence gates.From its SKILL.md
npx -y skills add yigityildiz0/universal-ai-skill-library --skill subagent-driven-developmentAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
2 things to look at
- no licenseNo license file was found in the repository. Code published without one is not open source by default, so using it at work is a question for whoever answers licensing questions where you are.
- 1 stars1 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
What its file declares
Copied from the file, not written here
The file declares its own license as MIT. That is the author’s claim about this one file, and it is not the same thing as the license GitHub reports for the repository, which is listed with the other numbers below.
SKILL.md
2.6 KB, 453 tokens by cl100k_base, as published. Nobody here has run it
Subagent-Driven Development
Entry gate
Use subagents when work can be isolated by dependency or an independent review is valuable. Keep a tightly coupled, small, or highly interactive change with the coordinator.
Task contract
Every delegated task receives:
Outcome and acceptance criteria:
Required source paths/context:
Dependencies already complete:
Exclusive write scope, or READ ONLY:
Forbidden files/actions:
Existing user changes to preserve:
Required tests/evidence:
Output and handoff format:
Stop/escalation conditions:
Use the host's actual delegation interface and current model/session defaults. Do not invent tools, force a named model, switch providers, or spend externally. Capability, context isolation, and tool access determine role assignment.
Execution
- Parse the approved plan into a dependency graph.
- Launch only dependency-ready tasks with non-overlapping mutable state.
- The coordinator continues non-overlapping integration work; it does not duplicate each implementation.
- Inspect returned files/diffs and run the claimed focused validation before accepting a task.
- Assign a specification reviewer read-only access to verify acceptance criteria from source evidence.
- Assign a quality/security reviewer when risk warrants it; do not leak the implementer's desired conclusion.
- Send confirmed findings back as a narrow fix contract. Re-review the resulting diff.
- Integrate in dependency order and run the complete relevant validation on the combined state.
Failure handling
NEEDS_CONTEXT: supply the missing source or narrow the contract.BLOCKED: verify the blocker, adjust scope/dependencies, or handle locally; do not automatically switch models.- overlapping edits: stop parallel writes and reconcile sequentially.
- failed validation: isolate the responsible integration and repair or roll it back.
- unavailable agent: continue locally or reassign the same bounded contract.
The parent task's authority is the ceiling. Subagents may not push, publish, deploy, delete unrelated data, spend, or message anyone unless that action was explicitly authorized.
Completion report
List the task graph, agent roles, accepted/rejected artifacts, independent checks, conflicts, commands/results, skipped validation, residual risk, and rollback. Completion requires integrated tests, not merely successful agent messages.
What ships with it
Read from the repository
Just SKILL.md. No reference files, no scripts.