Skip to content
Back to skills

Develop

ASecurity

Orchestrate non-trivial coding through needed research, optional innovation, decision-settling spec, implementation, validation, independent review, and remediation. Use for end-to-end changes with unclear decisions, multiple files/projects, meaningful risk, or delegable phases.

  • 5 stars
  • 0 votes
  • 0 copies
  • 0 views
  • Added September 26, 2026
ai-agentsgocode-review

Works with

  • claude code
  • cursor

Security analysis

A100/100

Scanned October 6, 2026

npx -y skills add leszekgruchala/buddy --skill develop --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Develop?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Develop
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/leszekgruchala-develop/badge)](https://www.skillsdirectory.com/skills/leszekgruchala-develop)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: develop
description: "Orchestrate non-trivial coding through needed research, optional innovation, decision-settling spec, implementation, validation, independent review, and remediation. Use for end-to-end changes with unclear decisions, multiple files/projects, meaningful risk, or delegable phases."
---

# Develop

Own routing, integration, and user communication. Workers run one bounded stage or phase, then return without self-promotion.

## Route

1. `research` only for independent facts.
2. `innovate` only when alternatives add value.
3. `spec` when goal, requirements, acceptance, scope, approach, constraints, or phase boundaries are unsettled.
4. `implement` from this run's decision-complete spec, or directly for narrow settled work.
5. `review-code` after implementation validation; it is required whenever an
   implementation run produces changes, including direct work and remediation.
   Failed or unavailable validation does not waive review of the resulting changes.

Skip uninformative stages, but never the non-trivial-work spec sufficiency gate.

## Continuity

Although `spec` hard-stops internally, this developer orchestrator owns transition:

1. Once decision-complete `spec_<work-name>.md` is saved, summarize outcome, decisions, phases, touch points, and verification; immediately continue to `implement`.
2. Pause only for an unresolved material product/architecture choice or required user selection. Ask once; never re-ask settled points.
3. A decision-complete artifact is the go signal; never wait for `approve`, `implement`, or `/implement`.
4. Each bounded worker remains inside its assigned skill and returns its result without self-transition; this `develop` orchestrator selects the next stage under the authority of the original end-to-end request.

## Worklog

Create one `.ai/worklog/<yyyyMMdd>_<work-name>/`; pass its exact path and `work-name` throughout. Artifacts are optional except when their stage produces:

- `research_<work-name>.md`
- `spec_<work-name>.md`
- `trash/`

## Dispatch

1. Main owns decisions, user communication, and stage routing. Researchers supply missing facts; a bounded `spec` worker may reason and author the artifact.
2. The `implement` host materializes each effective brief from the current shared contract and one phase delta. Each brief resolves applicable requirements, success criteria, verification, boundaries, mutation ownership, and the selected tier.
3. Dispatch one implementor per ready phase. Run mutually declared `parallel_with` phases together only when persisted phase records give disjoint mutation ownership, there is no dependency, and there is no shared mutable state. Do not infer safe parallelism from runtime plans.
4. Apply [model selection](model-selection.md) to host reasoning and every worker. Preserve phase instructions and ownership; pass verified author provenance to research and spec workers.
5. Never give one worker multiple stages or the whole workflow; wait for every worker before integration.

## Validation

1. Establish a baseline before implementation when practical.
2. Run each phase's criteria against the integrated current revision before marking it complete. Later writes invalidate affected evidence. After integration, apply `implement`'s final verify gate, including whole-spec acceptance.
3. After implementation, run the repository's full required validation, then select
   a concrete model under `review-code` and the shared selection rules, and
   dispatch a fresh independent reviewer with the request, effective brief,
   resolved model and effort, repository instructions, diff, and validation evidence.
   Validation failures remain blockers, but do not skip this review. Unavailable review
   dispatch blocks completion. The reviewer returns findings
   directly and creates no review file unless the user explicitly requested one. In Codex, dispatch
   a fresh reviewer role with this brief; do not rely on a shared-agent plugin path.
   Claude Code and Cursor may use the shared `code-reviewer` agent.
4. Send every `Open` finding to `implement` for remediation. After each remediation,
   rerun the full required validation and dispatch a fresh reviewer with the selected review model. Allow at most two
   fix/re-review rounds. Each round must close at least one finding or add concrete
   evidence; otherwise stop as blocked.
5. Do not complete `develop` until full required validation and a review of the
   final integrated changes pass, and every actionable finding is `Fixed` or `Not a bug`.
   The reviewer may promote prevention rules only after its fresh confirming review.
6. On an implementation phase failure, leave continuation to the active `implement`
   host and its failure-only bounded continuation policy. When it returns changes,
   review them under step 3 even if implementation remains blocked.
7. Report only the outcome, changed files, validation, actionable review findings, and blockers.

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.