recursive-mode skill
recursive-mode moves an agent task through file-backed phases, requirements to memory, and keeps the work trail in the repo.
- 136 ★ on the repository
- Last commit
- Apache-2.0
Extended access
Why this level
- The agent creates documents and branches in the repository and may call external CLIs through the router.
- Work delegated by subskills must be re-verified.
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add skills/recursive-modeDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- The agent creates documents and branches in the repository and may call external CLIs through the router.
- Work delegated by subskills must be re-verified.
This is third-party code. Review the repository files before installing.
What it does
recursive-mode is a skill set for structured agent-assisted development. It moves a task through phases: requirements, analysis, plan, implementation, verification, human or agent-operated QA and closeout. Each phase writes a document into `/.recursive/run/<run-id>/`, passes a draft, audit, repair and re-audit loop and is locked. Later corrections go in as addenda instead of rewriting locked history. The package bundles subskills: `recursive-spec` writes requirements, `recursive-worktree` prepares an isolated git worktree, `recursive-tdd` enforces RED-GREEN-REFACTOR, `recursive-debugging` finds root causes, `recursive-review-bundle` builds delegated-review material, `recursive-subagent` and `recursive-router` delegate bounded work, and `recursive-training` moves lessons from finished runs into `/.recursive/memory/`.
Who it is for. Developers and teams running tasks through an agent in a git repository who want an auditable trail: requirements, plan and verification evidence in files.
Good fit when
- An agent takes on a large feature and requirements must be pinned down before the first line of code.
- You must show that every requirement was implemented and verified.
- Work happens on a separate branch and worktree isolation should be part of the process.
- Finished runs pile up lessons worth feeding back into the next tasks.
- Review is delegated to a second agent while the main agent re-verifies the result.
Not a fit when
- The task is a one-line fix and five audited phases would only slow it down.
- The repository must not contain a service folder `/.recursive/`.
- You want a process without locks and repeated audits.
Example request
Draft requirements for the feature with recursive-spec, then implement the run in a separate worktree and record verification evidence.Limitations
The last commit in the repository is dated July 2026. The process is heavy: every phase creates documents and passes an audit, so small tasks cost more tokens and time. The `recursive-benchmark` add-on is not part of the default set. Routing through `recursive-router` starts only on explicit request and calls external CLIs with your settings.
How to disable. Remove the installed subskills from the agent's skills directory. Delete the `/.recursive/` folder in the repo separately if the run history is not needed.
FAQ
How do I install every subskill?
`npx skills add try-works/recursive-mode --skill '*' --full-depth`. A single subskill uses the same form with `--skill recursive-spec --full-depth`.
What is in the default set?
Nine skills: the main `recursive-mode`, `recursive-spec`, `recursive-worktree`, `recursive-debugging`, `recursive-tdd`, `recursive-review-bundle`, `recursive-router`, `recursive-subagent` and `recursive-training`.
Where do I start after installing?
Open a repository and, if requirements do not exist yet, ask for a plan or spec so `recursive-spec` triggers. Then give a short command such as `Implement the run`.
What does locking a phase mean?
After the audit passes, the phase document is closed. Later understanding is recorded in addenda and locked history is not rewritten.
Where does accumulated memory live?
In `/.recursive/memory/`, with the index in `MEMORY.md`. `recursive-training` fills it from finished runs.
Discussion
No questions or reviews yet.
Related
SkillsAn autonomous iteration skill for agents: modify, verify, revert failed changes, and repeat the loop toward a target metric
Eleven skills take an agent from an idea to verified code through a PRD and RFCs kept as files in the repository
Nine JS Mastery skills take a change through phases: from idea and spec to verification, tests, release texts and a docs sync
A Grid Dynamics plugin that loads a team's shared engineering standards, architecture and workflows into a coding agent session
A generation-based development pipeline: the agent studies the project, plans, codes, validates and logs lessons into knowledge.
ub-mode is a skill for Claude Code, Cursor and Codex that sets edit order: make it work first, then clean up by metrics.
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Repository stars | Main difference |
|---|---|---|---|
| recursive-mode (this card) | no | 136 | |
| OpenSpec | no | 74,784 | drives changes through specs; recursive-mode adds audited locked phases and run memory |
| BMad Method | no | 55,338 | builds the process on agent roles; recursive-mode builds it on file-backed phases and worktree, TDD and review subskills |
README in short
Requirements, plan and outcomes are kept as documents in the repository. The main agent verifies delegated work itself, and closeout results shape later runs.