New: 98,000+ skills indexed from GitHub repos with 100+ stars · every skill safety-checkedBrowse top skillsTop skills
Esc to closeBrowse all skills
Submit a skillBrowse skills
Submit a skillBrowse skills

Agent skill

Recursive Training

by try-works in try-works/recursive-mode

Use after completed recursive-mode runs accumulate to extract durable experiential memory into /.recursive/memory/, then load it for later runs through the canonical loader.

Apache-2.0Auto-check passedDevelopment

Install Recursive Training

skills CLI
$ npx skills add try-works/recursive-mode --skill recursive-training -a claude-code

Project install by default; add -g for ~/.claude/skills/.

GitHub CLI
$ gh skill install try-works/recursive-mode recursive-training --agent claude-code

Project scope by default; add --scope user for a personal install. Needs GitHub CLI 2.90.0 or later (public preview).

Manual copy
$ git clone --depth 1 https://github.com/try-works/recursive-mode.git skills-src && mkdir -p .claude/skills && cp -r skills-src/skills/recursive-training .claude/skills/recursive-training && rm -rf skills-src

Use ~/.claude/skills/ instead of .claude/skills for a personal install. The folder must contain SKILL.md.

Claude Code skills documentation · loads skills from .claude/skills/

Facts

Skill name
recursive-training
GitHub stars
136
Token cost
~1.9k tokens
SKILL.md length
710 words
Files
4 (incl. references)
Skills in repo
9
Repo updated
First seen
Licence
Apache-2.0

At a glance

Use after completed recursive-mode runs accumulate to extract durable experiential memory into /.recursive/memory/, then load it for later runs through the canonical loader.

  • Works in 8 steps: Repository-local only. Training data and… → No parameter updates. Learning happens… → /.recursive/memory/ is the only… → …
  • Development work in your project
  • SKILL.md covers Purpose, When to use it, Hard rules and Training model, plus 8 more sections
  • Calls python

What it does

Recursive Training is an agent skill from try-works/recursive-mode. Use after completed recursive-mode runs accumulate to extract durable experiential memory into /.recursive/memory/, then load it for later runs through the canonical loader.

Its SKILL.md is about 1.9k tokens, which your agent loads only when the skill is triggered. The skill folder holds 5 other files, including reference files (for example `agents/openai.yaml`, `references/memory-architecture.md` and `references/phase8-and-loading.md`).

It sits in Development. The repository describes itself as: Recursive workflow for agentic engineering. Like Factory Missions but properly recursive, free and open source. The licence is Apache-2.0.

When your agent uses it

  • Development work in your project

Example prompts

  • “/recursive-training”

Requirements

  • Python 3

Workflow steps

8 steps, taken from the first numbered list in SKILL.md.

  1. Repository-local only. Training data and extracted memory stay inside the current repo.
  2. No parameter updates. Learning happens through files in /.recursive/memory/, not model mutation.
  3. /.recursive/memory/ is the only canonical store. Pointer files are bootstrap-managed and non-authoritative.
  4. All markdown under /.recursive/run// is eligible training input, not just 00-08.
  5. Group runs by subsystem only. The extractor assigns task types per learning item.
  6. Use contrastive extraction when both winners and losers exist; fall back to winner-only extraction for high-quality repos.
  7. Every extracted item must remain evidence-grounded in completed runs.
  8. Training scripts do not own AGENTS.md; bootstrap owns bridge-file updates.

What it can do on your machine

Read from SKILL.md and the folder at commit ff75bc7. It shows what the files ask for, not the result of running them.

  • Tool permissions

    Pre-approves nothing: there is no allowed-tools line, so your agent's usual permission prompts apply.

    From allowed-tools in the SKILL.md frontmatter.

  • Runs code

    Shell commands in SKILL.md call:

    • python

    From the folder's file list and the shell code blocks in SKILL.md.

  • Network

    No URLs in SKILL.md.

    From URLs in SKILL.md, links to its own repository left out.

  • Credentials

    Names no API keys, tokens, secrets or passwords.

    From names ending in _API_KEY, _TOKEN, _SECRET, _KEY or _PASSWORD in SKILL.md.

Context cost

Recursive Training loads about 1.9k tokens when it runs, and up to ~4.1k if it reads all its reference files. Until then it costs about 49 tokens; SKILL.md has 710 words of instructions outside code blocks.

Always · name and description, kept in context so the agent knows when to use it
~49
When it runs · the whole SKILL.md, loaded when a task matches
~1.9k
With references · SKILL.md plus every file in references/, read only if the agent opens them
~4.1k

Estimates: characters ÷ 4, the usual rule of thumb; real counts depend on the model's tokenizer. Scripts and assets cost tokens only if the agent reads them.

Safety

Auto-check passed

The automated check found no risky patterns in SKILL.md.

Automated static check — not a guarantee. Review scripts before installing. It scans the text of SKILL.md for risky patterns (piping downloads into a shell, reading credential files, hidden Unicode, destructive commands); files beside SKILL.md are not scanned.

SKILL.md

The full file from try-works/recursive-mode at commit ff75bc7, republished under its Apache-2.0 licence (© try-works). 710 words, ~1,853 tokens.

Download SKILL.mdSave it as .claude/skills/recursive-training/SKILL.md (or your agent's skills folder). This skill also uses 3 other files; get the full folder from GitHub.
name
recursive-training
description
Use after completed recursive-mode runs accumulate to extract durable experiential memory into `/.recursive/memory/`, then load it for later runs through the canonical loader.

recursive-training

Purpose

Use this skill after a repository has accumulated completed recursive-mode runs and you want to turn repeated successes or failures into durable, repo-local guidance.

The canonical workflow still lives in /.recursive/RECURSIVE.md. This skill owns only the training, loading, and memory-discipline layer that sits around completed runs.

When to use it

Use recursive-training to:

  • extract cross-run learnings from completed recursive-mode runs
  • keep those learnings in /.recursive/memory/ instead of ad hoc mirrors
  • refresh memory after Phase 8 locks
  • load only the most relevant prior learnings before a new run starts
  • provide startup guidance without mutating the memory plane

Hard rules

  1. Repository-local only. Training data and extracted memory stay inside the current repo.
  2. No parameter updates. Learning happens through files in /.recursive/memory/, not model mutation.
  3. /.recursive/memory/ is the only canonical store. Pointer files are bootstrap-managed and non-authoritative.
  4. All markdown under /.recursive/run/<run-id>/ is eligible training input, not just 00-08.
  5. Group runs by subsystem only. The extractor assigns task types per learning item.
  6. Use contrastive extraction when both winners and losers exist; fall back to winner-only extraction for high-quality repos.
  7. Every extracted item must remain evidence-grounded in completed runs.
  8. Training scripts do not own AGENTS.md; bootstrap owns bridge-file updates.

Training model

recursive-training combines two ideas:

  • ReasoningBank-style memory items for structured, reusable extracted guidance
  • Training-free GRPO-style comparison for contrastive winner/loser extraction when variance exists

At a high level:

  1. Parse all markdown artifacts from completed runs.
  2. Infer the dominant subsystem from changed paths and evidence across those artifacts.
  3. Classify each subsystem group as contrastive, winner-only, or insufficient.
  4. Ask the extractor for structured learning items.
  5. Write those items into:
    • /.recursive/memory/domains/<subsystem>.md
    • /.recursive/memory/training/<task-type>.md
  6. Refresh the memory registry/startup guidance without treating pointer files as authoritative memory.

For the full schema, grouping logic, evidence signals, and extraction contracts, see:

  • references/memory-architecture.md
  • references/phase8-and-loading.md

Phase 8 and loading boundary

Phase 8 records the current run's observations in 08-memory-impact.md.

Training then turns many completed runs into cross-run memory.

That means:

  • 08-memory-impact.md is run-local capture
  • recursive-training-phase8-trigger.py is the handoff after Phase 8 locks
  • recursive-training-grpo.py builds prompts and writes memory after extraction
  • recursive-training-extract.py evaluates prompts (required companion; not optional)
  • recursive-training-loader.py is the canonical retrieval path before later runs
  • recursive-training-sync.py prints startup guidance without mutating the memory plane

Extractor wiring (first match wins):

  1. --response-file / sibling <prompt>.response.json for agent-operated offline evaluation
  2. RECURSIVE_TRAINING_EXTRACTOR_CMD with {prompt_file} / {repo_root} placeholders
  3. otherwise exit 2 (unavailable) — GRPO/trigger must fail, not claim success

Detailed Phase 8 handoff and loader behavior lives in references/phase8-and-loading.md.

Show full SKILL.md (295 more words)Show less

Commands

Full training:

bash
python .recursive/scripts/recursive-training-grpo.py --repo-root .

Incremental training after a specific run:

bash
python .recursive/scripts/recursive-training-grpo.py --repo-root . --incremental --run-id <run-id>

Post-Phase 8 trigger:

bash
python .recursive/scripts/recursive-training-phase8-trigger.py --repo-root . --run-id <run-id>
python .recursive/scripts/recursive-training-phase8-trigger.py --repo-root . --run-id <run-id> --auto

Read-only startup guidance:

bash
python .recursive/scripts/recursive-training-sync.py --repo-root .

Canonical memory loading before a new run:

bash
python .recursive/scripts/recursive-training-loader.py \
  --repo-root . \
  --query "<task description>" \
  --files "<comma-separated paths>"

Optional MCP convenience layer:

bash
python .recursive/scripts/recursive-training-mcp.py --repo-root .

Trigger patterns

Recognize these as training requests:

  • train
  • training
  • extract memories
  • learn from runs
  • train from the latest run
  • incremental training
  • sync memories
  • /recursive-training

Default behavior:

  • fewer than 2 completed runs: explain why extraction is skipped
  • no explicit scope: default to full training
  • explicit run or "incremental": use incremental mode
  • "sync" or "what should I read": use recursive-training-sync.py

Operator checklist

  1. Confirm the repo already has recursive-mode scaffolding.
  2. Confirm completed runs exist under /.recursive/run/.
  3. Run the trigger or grpo script with the intended scope.
  4. Verify updated items land under /.recursive/memory/domains/ and /.recursive/memory/training/.
  5. Before the next run, read /.recursive/memory/MEMORY.md and call the loader with task context.
  6. Treat loader output as advisory context; the canonical records stay in the memory plane.

Script surface

  • .recursive/scripts/recursive-training-grpo.py
  • .recursive/scripts/recursive-training-grpo.ps1
  • .recursive/scripts/recursive-training-extract.py
  • .recursive/scripts/recursive-training-extract.ps1
  • .recursive/scripts/recursive-training-phase8-trigger.py
  • .recursive/scripts/recursive-training-phase8-trigger.ps1
  • .recursive/scripts/recursive-training-sync.py
  • .recursive/scripts/recursive-training-sync.ps1
  • .recursive/scripts/recursive-training-loader.py
  • .recursive/scripts/recursive-training-loader.ps1
  • .recursive/scripts/recursive-training-mcp.py
  • .recursive/scripts/recursive-training-mcp.ps1

Detailed references

  • references/memory-architecture.md — schema, evidence signals, grouping, extraction modes, and output contract
  • references/phase8-and-loading.md — Phase 8 handoff, trigger behavior, loader timing, and integration patterns

Coverage Gate

  • Canonical memory-store boundary documented
  • Contrastive and winner-only extraction modes documented
  • Subsystem-only grouping documented
  • All-markdown run-folder input scope documented
  • Phase 8 handoff and loader boundary documented
  • Canonical command surface documented
  • Reference docs linked for detailed extraction and loading behavior
  • Script-only transport boundary preserved

Coverage: PASS

Approval Gate

  • Skill defers to /.recursive/RECURSIVE.md for overall workflow rules
  • Repository-local scope preserved
  • No parameter-updating behavior introduced
  • /.recursive/memory/ remains the only canonical store
  • Phase 8 capture vs training extraction boundary is explicit
  • Loader remains the canonical retrieval path
  • Pointer files remain bootstrap-managed, not sync-authored mirrors

Approval: PASS

© try-works, Apache-2.0. Rendered from Markdown: HTML in the file is shown as text, images as links, and headings moved down two levels. Raw file

Files

SKILL.md and 3 other files (references) in skills/recursive-training of try-works/recursive-mode.

  • SKILL.md
  • agents/openai.yaml
  • references/memory-architecture.md
  • references/phase8-and-loading.md

Open the folder on GitHubat commit ff75bc7

Used in 1 other repository

We found 1 copy of this SKILL.md (exact, near-identical or edited) in other folders. This page covers the copy in try-works/recursive-mode, which our catalogue first saw on October 7, 2026.

Compare with similar skills

Recursive Training next to the 5 skills that share the most tags, products or categories with it. Stars are the repository's; “used in” counts other GitHub owners with a copy.

Recursive Training compared with similar skills
SkillStarsUsed inTokensAuto-checkLicenceRepo updated
Recursive Training this skilltry-works/recursive-mode136—~1.9kAutomated safety check: PassApache-2.0
Vercel Composition Patternssupabase/supabase111k59 repos~726Automated safety check: PassMIT
Finishing a Development Branchobra/superpowers296k5 repos~1.9kAutomated safety check: PassMIT
Typescript Advanced Typesrolling-scopes/rsschool-app10k25 repos~4.2kAutomated safety check: PassMPL-2.0
PR Babysitteropeninterpreter/openinterpreter69k3 repos~4.2kAutomated safety check: PassApache-2.0
Code Review ChecklistshareAI-lab/learn-claude-code78k5 repos~1.1kAutomated safety check: PassMIT

Similar skills

  • Vercel Composition Patterns

    supabase/supabase

    Official

    React composition patterns that scale. An agent skill from supabase/supabase.

    111k GitHub starsUsed in 59 repos~726 tokens
    DevelopmentAuto-check passed

  • Finishing a Development Branch

    obra/superpowers

    Walks the last step of a branch: confirm tests pass, detect the git environment, ask how to integrate, carry out your choice and clean up the worktree.

    296k GitHub starsUsed in 5 repos~1.9k tokens
    DevelopmentAuto-check passed

  • Typescript Advanced Types

    rolling-scopes/rsschool-app

    Master TypeScript's advanced type system including generics, conditional types, mapped types, template literals, and utility types for building type-safe applications.

    10k GitHub starsUsed in 25 repos~4.2k tokens
    DevelopmentAuto-check passed

  • PR Babysitter

    openinterpreter/openinterpreter

    Watches an open GitHub pull request until it merges, handling review comments, diagnosing CI failures and retrying flaky checks along the way.

    69k GitHub starsUsed in 3 repos~4.2k tokens
    DevelopmentAuto-check passed

  • Code Review Checklist

    shareAI-lab/learn-claude-code

    Reviews code against a five-part checklist covering security, correctness, performance, maintainability and testing, and reports findings in a fixed format.

    78k GitHub starsUsed in 5 repos~1.1k tokens
    DevelopmentAuto-check passed

  • Greploop

    onyx-dot-app/onyx

    Iteratively improves a PR (GitHub), MR (GitLab), or shelved changelist (Perforce) until Greptile gives it a 5/5 confidence score with zero unresolved comments.

    32k GitHub starsUsed in 4 repos~3.3k tokens
    DevelopmentAuto-check passed

More from try-works/recursive-mode

All 9 skills in this repo

  • Recursive Mode

    try-works/recursive-mode

    Repository workflow orchestration skill for staged implementation, locked artifacts, late-phase receipts, and durable memory maintenance.

    136 GitHub stars~3.4k tokensUpdated 2 mo ago
    Auto-check passed

  • Recursive Debugging

    try-works/recursive-mode

    A skill your agent uses when a recursive-mode requirement involves debugging a bug, test failure, or unexpected behavior.

    136 GitHub stars~3.4k tokensUpdated 2 mo ago
    Auto-check passed

  • Recursive Review Bundle

    try-works/recursive-mode

    A skill your agent uses when recursive-mode work needs a canonical delegated-review or audit handoff.

    136 GitHub stars~873 tokensUpdated 2 mo ago
    Auto-check passed

  • Recursive Subagent

    try-works/recursive-mode

    A skill your agent uses when recursive-mode work may benefit from delegated audit, review, or bounded implementation support.

    136 GitHub stars~3k tokensUpdated 2 mo ago
    Auto-check passed

  • Recursive TDD

    try-works/recursive-mode

    A skill your agent uses when implementing any code in recursive-mode Phase 3.

    136 GitHub stars~2.8k tokensUpdated 2 mo ago
    Auto-check passed

  • Recursive Worktree

    try-works/recursive-mode

    A skill your agent uses when starting any recursive-mode requirement to set up an isolated git worktree.

    136 GitHub stars~1.2k tokensUpdated 2 mo ago
    Auto-check passed

Categories

Questions about Recursive Training

What does Recursive Training do?

Use after completed recursive-mode runs accumulate to extract durable experiential memory into /.recursive/memory/, then load it for later runs through the canonical loader. Recursive Training is an agent skill from try-works/recursive-mode.recursive/memory/, then load it for later runs through the canonical loader.

When should I use Recursive Training?

Recursive Training fits situations like: development work in your project.

How do I install Recursive Training in Claude Code?

Run `npx skills add try-works/recursive-mode --skill recursive-training -a claude-code`. Or copy the skill folder (skills/recursive-training in try-works/recursive-mode) into .claude/skills/recursive-training in your project. Claude Code loads it when a task matches its description.

How do I install Recursive Training in Codex?

Run `npx skills add try-works/recursive-mode --skill recursive-training -a codex`. Or copy the skill folder (skills/recursive-training in try-works/recursive-mode) into .agents/skills/recursive-training in your project. Codex loads it when a task matches its description.

Can I use Recursive Training in Cursor, Gemini CLI or GitHub Copilot?

Cursor, Gemini CLI, GitHub Copilot and OpenCode also load SKILL.md folders. With the skills CLI, run `npx skills add try-works/recursive-mode --skill recursive-training -a cursor` (or -a gemini-cli, github-copilot or opencode for the others). To copy it by hand, put the folder in .cursor/skills/recursive-training, .gemini/skills/recursive-training, .github/skills/recursive-training and .opencode/skills/recursive-training in your project.

What does Recursive Training need to run?

Going by SKILL.md and its folder, Recursive Training needs the command-line tools its instructions call (python). Our summary lists: Python 3.

Does Recursive Training access the network?

SKILL.md contains no URLs. Any network use would come from the scripts or tools the agent runs. This is read from the text; nothing was executed.

Is Recursive Training safe to install?

Our automated static check of SKILL.md found no risky patterns, such as piping downloads into a shell, reading credential files or hidden Unicode. It is not a guarantee. Review the folder before installing.

What licence does Recursive Training use?

Recursive Training is published under the Apache-2.0 licence (the repository's licence). It allows redistribution, so the full SKILL.md is shown on this page.

How many tokens does Recursive Training use?

About 1.9k tokens (SKILL.md is roughly 7.4k characters). Agents keep only the skill's name and description in context until a task matches; then they load SKILL.md in full. Its references folder adds about 2.2k tokens, read only when the agent opens those files.

What are the alternatives to Recursive Training?

Skills that share tags, products or a category with Recursive Training: Vercel Composition Patterns (supabase/supabase, 111k stars), Finishing a Development Branch (obra/superpowers, 296k stars), Typescript Advanced Types (rolling-scopes/rsschool-app, 10k stars) and PR Babysitter (openinterpreter/openinterpreter, 69k stars). The comparison table on this page puts their stars, adoption, token cost, safety result and licence side by side.

Who maintains Recursive Training?

try-works (a GitHub user) maintains it in try-works/recursive-mode, which has 136 GitHub stars. The repository holds 9 skills in this directory. The repository was last updated on July 24, 2026.

Source: try-works/recursive-mode on GitHub. Facts on this page come from the repository at the commit we read; the author's words are quoted as theirs.