Build, continue, inspect, or visualize a source-grounded mathematical roadmap and theorem DAG in an existing Autoform Markdown blueprint.

MITAuto-check passedProduct & Project Management

Install Roadmap

skills CLI
$ npx skills add facebookresearch/autoform-bot --skill roadmap -a claude-code

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

GitHub CLI
$ gh skill install facebookresearch/autoform-bot roadmap --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/facebookresearch/autoform-bot.git skills-src && mkdir -p .claude/skills && cp -r skills-src/skills/roadmap .claude/skills/roadmap && 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
roadmap
GitHub stars
117
Token cost
~1.8k tokens
SKILL.md length
960 words
Files
3 (incl. references)
Skills in repo
6
Repo updated
First seen
Licence
MIT

At a glance

Build, continue, inspect, or visualize a source-grounded mathematical roadmap and theorem DAG in an existing Autoform Markdown blueprint.

  • Source research
  • SKILL.md covers Own the whole pass, Ground and decompose and Finish
  • Instructions only: no scripts, shell commands, URLs or credentials in SKILL.md
  • Roadmap articles

What it does

Roadmap is an agent skill from facebookresearch/autoform-bot. Build, continue, inspect, or visualize a source-grounded mathematical roadmap and theorem DAG in an existing Autoform Markdown blueprint. Use for source research, scope, coverage, milestones, and roadmap articles; do not create repository infrastructure or prove Lean declarations.

Its SKILL.md is about 1.8k tokens, which your agent loads only when the skill is triggered. The skill folder holds 4 other files, including reference files (for example `agents/openai.yaml` and `references/cabannes-thesis-roadmap.md`).

It sits in Product & Project Management, covering Project management and Source-grounded notebooks. The licence is MIT.

When your agent uses it

  • Source research
  • Roadmap articles
  • Do not create repository infrastructure
  • Prove Lean declarations

Example prompts

  • “/roadmap”

What it can do on your machine

Read from SKILL.md and the folder at commit 4c80749. 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

    No scripts in the folder and no shell commands in SKILL.md.

    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

Roadmap loads about 1.8k tokens when it runs, and up to ~2.5k if it reads all its reference files. Until then it costs about 72 tokens; SKILL.md has 960 words of instructions outside code blocks.

Always · name and description, kept in context so the agent knows when to use it
~72
When it runs · the whole SKILL.md, loaded when a task matches
~1.8k
With references · SKILL.md plus every file in references/, read only if the agent opens them
~2.5k

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 facebookresearch/autoform-bot at commit 4c80749, republished under its MIT licence (© facebookresearch). 960 words, ~1,771 tokens.

Download SKILL.mdSave it as .claude/skills/roadmap/SKILL.md (or your agent's skills folder). This skill also uses 2 other files; get the full folder from GitHub.
name
roadmap
description
Build, continue, inspect, or visualize a source-grounded mathematical roadmap and theorem DAG in an existing Autoform Markdown blueprint. Use for source research, scope, coverage, milestones, and roadmap articles; do not create repository infrastructure or prove Lean declarations.

Build an Autoform roadmap

Maintain an ordered mathematical book whose formalizable leaves form a dependency DAG of coherent, pull-request-sized units. Markdown under blueprint/ is the source of truth. Consult the format and command reference only when exact syntax or CLI usage is needed, and the worked example only when a concrete source-to-DAG pattern would help.

Own the whole pass

Honor a request explicitly limited to inspection, audit, status, or visualization. Otherwise, a direct user invocation of Roadmap requests one complete planning pass over the source boundary named by the user or already adopted by the vault.

That direct invocation explicitly requests persistent execution. When the runtime exposes a model-callable Goal lifecycle, use it before other work: continue a compatible active Goal or create one for a complete, validated, reconciled, and committed roadmap of that source boundary. Do not set a token budget unless the user supplied one, and never replace an unrelated active Goal. Do not ask the user to invoke another command, shell out to a host's interactive goal command, or substitute todos or repository files for native continuation. If no compatible Goal can be used, complete the same pass in the current run and leave any unrelated Goal unchanged.

Treat source discovery, coarse coverage, chapter boundaries, fine decomposition, and validation as internal checkpoints. Do not pause for approval unless the user requested staged review. Make reversible choices and record assumptions. Ask one concise question only after inspection cannot identify or access the source, or when incompatible scope choices remain and choosing one would discard accepted work.

Ground and decompose

Inspect the repository and existing vault before writing. Preserve accepted material and unrelated changes; send missing infrastructure to Setup. Formalize owns Lean execution after this skill produces a validated ready frontier.

Work from exact source passages. Record stable source locations, assumptions, and uncovered prerequisites under blueprint/sources/; label project-authored specifications honestly. Search the pinned Mathlib checkout before planning a replacement. External research is read-only, and contacting people requires permission. Read Setup's Zulip workflow only for requested Zulip work.

Enumerate the entire adopted boundary in blueprint/coverage/README.md. MAPPED is unfinished, DECOMPOSED links to roadmap nodes, DEFERRED records an explicit user decision or concrete external blocker, and OUT explains an exclusion. Treat every other DEFERRED row as queued work; never defer merely to shorten the run. Because coverage.complete only checks declared rows, compare the table with the source structure yourself.

Write milestone pages under blueprint/roadmap/ by mathematical significance, then one fine article per coherent unit with one unique main result. Ground each statement and proof sketch in the source. Put genuine statement prerequisites under ## Depends on and proof-only prerequisites under ## Proof depends on. Assign durable article_id metadata to new articles; use autoform migrate article-ids blueprint --json to obtain deterministic IDs after creating the pages. Assert formalization or mathlib: true only after exact verification. Before revising a formalizable leaf, acquire the claim_target that autoform work context reports for it, passing your own --worker-id; renew it while editing and release it once the committed revision is on the branch Formalize works from. A refused acquire means another agent owns the article: leave it and report it. Claims write refs to the board's remote, which is outward-facing, so make sure the request covers them. When a revision changes a statement whose article has lean:, retract it: replace statement: formalized with statement: retracted, remove proof: formalized, and keep lean:, which Formalize needs to run autoform work impact; an article without lean: just loses statement and proof. When the project's CI pins an AUTOFORM_REF older than the marker, its autoform check rejects statement: retracted: remove statement and proof and keep lean: until the pin moves, and report the old pin. Under the open policy a retracted theorem stays an open statement, so whatever rests on it stays conditionally proved. Retract only that article and the dependents whose Markdown text the revision rewrites, claiming them all in one acquire; the Lean-side impact decides every other dependent. For a Lean revision requested in Human Review, record the decision in the article and retract it the same way, so it returns to the frontier as a statement phase flagged as a revision, and leave the Lean change to Formalize, which follows the revision contract; this skill edits only Markdown. For a large source, divide independent sections among available agents while retaining one owner for global coverage and dependency consistency.

Show full SKILL.md (232 more words)Show less

open_statements in roadmap/README.md is a project policy; absent means strict. allowed lets a theorem's statement land with a sorry proof, so dependents can be stated and proved before it is; the cost is conditionally proved results that stay incomplete until those proofs land. Change it only on the user's request and only once CI meets the open statements requirements. Turn it back off only once no open statement remains: the strict audit rejects every sorry, and strict status shows a proof resting on one as proved.

Reconcile every affected source and milestone page, the coverage contract, blueprint/README.md, and the repository README.md.

Finish

After the final edit, use the CLI reference to run autoform check and autoform audit, refresh the Mermaid graph, and resolve every finding introduced by this work or inside the adopted boundary. Report unrelated pre-existing findings instead of silently widening scope. Commit the vault and refreshed graph only after this final validation; pushing is outward-facing and requires a user request.

Finish only when the adopted boundary has no MAPPED rows, every DECOMPOSED area links to a source-grounded fine DAG, affected pages and the graph agree, and the latest commit contains every change from the pass. Do not stop after discovery, a coarse proposal, one chapter, or unchanged validation. Report the material delta, evidence, remaining explicit blockers, and next execution frontier. Mark an active Goal complete only after these conditions hold.

© facebookresearch, MIT. 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 2 other files (references) in skills/roadmap of facebookresearch/autoform-bot.

  • SKILL.md
  • agents/openai.yaml
  • references/cabannes-thesis-roadmap.md

Open the folder on GitHubat commit 4c80749

Compare with similar skills

Roadmap 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.

Roadmap compared with similar skills
SkillStarsUsed inTokensAuto-checkLicenceRepo updated
Roadmap this skillfacebookresearch/autoform-bot117—~1.8kAutomated safety check: PassMIT
Gsd Plant SeedSpillwaveSolutions/agent-brain1202 repos~691Automated safety check: PassMIT
Review Planmingchen666/Reviva236—~1.3kAutomated safety check: NotesNone
Project Managerespennilsen/pi122—~2.1kAutomated safety check: PassMIT
CCPM Project Managementautomazeio/ccpm8.4k—~1.1kAutomated safety check: PassMIT
Uvastral-sh/claude-code-plugins3142 repos~980Automated safety check: PassApache-2.0

Similar skills

  • Gsd Plant Seed

    SpillwaveSolutions/agent-brain

    Capture a forward-looking idea with trigger conditions — surfaces automatically at the right milestone

    120 GitHub starsUsed in 2 repos~691 tokens
    Product & Project ManagementAuto-check passed
  • Review Plan

    mingchen666/Reviva

    This skill should be used when the user asks to "做个复习计划", "安排复习", "考试还剩X天怎么安排", "制定学习计划", "帮我规划复习", "冲刺计划", "每天学什么", "复习时间怎么分配", "怎么安排复习", "帮我排个时间表", "期末复习怎么安排", "还有X天考试来得及吗", "帮我做个冲刺安排", or when…

    236 GitHub stars~1.3k tokensUpdated 16 days ago
    Product & Project ManagementAuto-check: notes
  • Project Manager

    espennilsen/pi

    Manage Espen's projects across the Dev folder and Obsidian vault.

    122 GitHub stars~2.1k tokensUpdated 16 days ago
    Product & Project ManagementAuto-check passed
  • Runs a spec-driven workflow from PRD to epic to GitHub issues to parallel agents, with status, standup and blocked-work reports from bundled scripts.

    8.4k GitHub stars~1.1k tokensUpdated 6 mo ago
    Product & Project ManagementAuto-check passed
  • Uv

    astral-sh/claude-code-plugins

    Official

    Guide for using uv, the Python package and project manager. An agent skill from astral-sh/claude-code-plugins.

    314 GitHub starsUsed in 2 repos~980 tokens
    Product & Project ManagementAuto-check passed
  • Project Management

    kunchenguid/firstmate

    Agent-only procedure for Firstmate project management. An agent skill from kunchenguid/firstmate.

    7.6k GitHub stars~2.1k tokensUpdated today
    Product & Project ManagementAuto-check passed

More from facebookresearch/autoform-bot

  • Setup

    facebookresearch/autoform-bot

    Set up, inspect, or repair repository infrastructure for an Autoform Lean project, including the Lean/Mathlib shell, an in-repository Obsidian-compatible blueprint vault, ignore rules, MkDocs…

    117 GitHub stars~2.5k tokensUpdated today
    Auto-check passed
  • Agent Review

    facebookresearch/autoform-bot

    Judge an Autoform mathematical roadmap or Lean formalization with explicit, evidence-based rubrics.

    117 GitHub stars~859 tokensUpdated today
    Auto-check passed
  • Develop Plugin

    facebookresearch/autoform-bot

    Maintain AutoformBot's code, skills, tests, examples, and installation.

    117 GitHub stars~417 tokensUpdated today
    Auto-check passed
  • Formalize

    facebookresearch/autoform-bot

    Formalize ready leaves from an existing Autoform Markdown roadmap in Lean, using native agents, fail-closed claims, and verified Markdown progress.

    117 GitHub stars~2.5k tokensUpdated today
    Auto-check passed
  • Human Review

    facebookresearch/autoform-bot

    Prepare and guide human inspection of an Autoform roadmap or formalization through its Obsidian graph and rendered blueprint site.

    117 GitHub stars~856 tokensUpdated today
    Auto-check passed

Questions about Roadmap

What does Roadmap do?

Build, continue, inspect, or visualize a source-grounded mathematical roadmap and theorem DAG in an existing Autoform Markdown blueprint. Roadmap is an agent skill from facebookresearch/autoform-bot. Build, continue, inspect, or visualize a source-grounded mathematical roadmap and theorem DAG in an existing Autoform Markdown blueprint.

When should I use Roadmap?

Roadmap fits situations like: source research; roadmap articles; do not create repository infrastructure; prove Lean declarations.

How do I install Roadmap in Claude Code?

Run `npx skills add facebookresearch/autoform-bot --skill roadmap -a claude-code`. Or copy the skill folder (skills/roadmap in facebookresearch/autoform-bot) into .claude/skills/roadmap in your project. Claude Code loads it when a task matches its description.

How do I install Roadmap in Codex?

Run `npx skills add facebookresearch/autoform-bot --skill roadmap -a codex`. Or copy the skill folder (skills/roadmap in facebookresearch/autoform-bot) into .agents/skills/roadmap in your project. Codex loads it when a task matches its description.

Can I use Roadmap 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 facebookresearch/autoform-bot --skill roadmap -a cursor` (or -a gemini-cli, github-copilot or opencode for the others). To copy it by hand, put the folder in .cursor/skills/roadmap, .gemini/skills/roadmap, .github/skills/roadmap and .opencode/skills/roadmap in your project.

What does Roadmap need to run?

SKILL.md names no scripts, command-line tools or credentials: Roadmap is instructions for the agent only.

Does Roadmap 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 Roadmap 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 Roadmap use?

Roadmap is published under the MIT licence (the repository's licence). It allows redistribution, so the full SKILL.md is shown on this page.

How many tokens does Roadmap use?

About 1.8k tokens (SKILL.md is roughly 7.1k 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 748 tokens, read only when the agent opens those files.

What are the alternatives to Roadmap?

Skills that share tags, products or a category with Roadmap: Gsd Plant Seed (SpillwaveSolutions/agent-brain, 120 stars), Review Plan (mingchen666/Reviva, 236 stars), Project Manager (espennilsen/pi, 122 stars) and CCPM Project Management (automazeio/ccpm, 8.4k stars). The comparison table on this page puts their stars, adoption, token cost, safety result and licence side by side.

Who maintains Roadmap?

facebookresearch (a GitHub organization) maintains it in facebookresearch/autoform-bot, which has 117 GitHub stars. The repository holds 6 skills in this directory. The repository was last updated on October 7, 2026.

Source: facebookresearch/autoform-bot on GitHub. Facts on this page come from the repository at the commit we read; the author's words are quoted as theirs.