supermodo
UserA coherent, opinionated toolkit for documentation-driven development with twin-model adversarial checks.
Categories
Indexed Skills (16)
flow
Runs the whole supermodo development pipeline for one task, each stage in its own subagent so the main context stays small. Use when the user wants to run the whole pipeline, "take this task from spec to commit", orchestrate a feature end-to-end, run the full flow (or the flow from a given stage), or drive a task through the complete dev-to-commit process — anything that means coordinating grill → work → hunt → tdd → tests → refactor → librarian → commit rather than a single stage.
sync-configs
Audit and sync Claude Code and Codex CLI configurations so both tools share the same instructions, skills, subagents, MCP servers, hooks, and permission intent. Accepts an optional argument naming one skill, agent, or config file to sync just that item without a full audit. Use this whenever the user mentions config drift, syncing/aligning Claude and Codex, AGENTS.md vs CLAUDE.md differences, keeping skills or agents consistent across AI CLIs, or after editing one tool's config and wanting the other to match — e.g. "sync this skill to codex", "I changed my-agent.md, mirror it". Also use when the user asks "are my Claude and Codex setups aligned?" or wants to migrate config from one tool to the other.
tests
Fixes failing tests and lint/type errors, audits test-suite quality, or drives coverage — for any project configured with a supermodo skills.config.json. Use whenever the user mentions failing tests, lint errors, "make tests pass", green build, flaky tests, test quality, weak or missing tests, corner cases, coverage, mutation testing, or asks whether the tests actually protect the core logic — even without "/tests".
grill
Twin-agent adversarial interview that locks a task or design before implementation. Two models plan independently, attack each other's plans, and only genuinely user-owned questions reach you — the locked outcome is handed to librarian to write. Use for 'grill', 'grill me', plan stress-test, design interview, new-task intake, spec lock, or a flow `needs-input` escalation.
protocols
The supermodo package's single source of truth and help desk. Holds the master copies of every shared protocol (config contract, docs convention, reports, subagent handoff, cross-model adversary ops, adversarial review stance, user questions, grilling, command tooling) that all other supermodo skills read in place. Invoke to ask how the supermodo package or any of its protocols works, to get an overview of the skills and pipeline, or to check that a supermodo installation is complete and healthy — triggers on "how does supermodo work", "explain the grilling/questions/docs protocol", "supermodo help", "check my supermodo install".
commit
Generate a one-line Conventional Commits message from the current git changes, then offer to run the commit for the user. Use whenever the user asks for a commit message or wants to commit finished work — "commit message", "write a commit", "generate commit", "commit msg", "cm", "/commit" — even when they don't name this skill. Reads the actual diff to derive what was done. Prints the message; then shows the exact commands and asks whether to commit — runs git only on an explicit yes or a configured auto-consent policy. Works in fresh folders too: no repository yet → same flow, with git init prepended to the offered commands (init always asks, even in auto mode).
config
Keeper of a project's supermodo configuration (skills.config.json) and docs scaffold. Use whenever the user wants to view, check, change, fix, or upgrade the supermodo config, set up or onboard a project onto the supermodo skills, start a new supermodo project, or whenever another supermodo skill reports a missing or invalid skills.config.json. Trigger on 'supermodo config', 'set up supermodo', 'configure supermodo', 'bootstrap supermodo', 'change the coverage target', 'edit the config', 'add a command to the config' — even when the word 'config' itself is not used.
hunt
Systematic bug hunting across a full-stack TypeScript application. Finds semantic bugs, async issues, data integrity problems, security vulnerabilities, performance hotspots, and UI/browser issues — producing a prioritized report with evidence and fix suggestions. Use when the user says /hunt, asks to "find bugs", "audit this", "check for issues", "scan for problems", wants a code review focused on correctness rather than style, mentions specific concerns like "are there race conditions", "check for security issues", "find performance problems", or wants to validate code quality before a release. Also triggers on "what's wrong with this", "anything broken", "smell check", or any request to systematically find problems in code.
librarian
Sole owner of documentation mutations: run the lifecycle pass to close out completed work, reconcile docs and agent definitions with code, promote verified contracts, record ADRs, archive finished tasks, regenerate navigation, and repair links. Also manages the backlog, runs new-task intake, promotes hunt or test-audit findings into work items (--promote), and (with --absorb) performs the one-time sweep that classifies and absorbs pre-existing documentation into the convention. Use for docs maintenance, task closeout, documentation or instruction drift, backlog operations, `--task` intake, `--promote` of findings into triads, `--absorb` onboarding of existing docs, or a flow stage-7 alignment pass.
refactor
Deep functional refactoring of TypeScript code for purity, testability, and modularity. Use when the user says /refactor, asks to clean up or restructure code, mentions large files, god functions, mixed concerns, poor testability, or when code grew unwieldy after feature iterations. Also triggers on "this file is too big", "split this", "make this more functional", "extract", "decompose", or any mention of code smells like let, try/catch in business logic, for loops, or files over 300 lines. If the user describes messy, tangled, or overgrown code that needs structural improvement, this is the right skill — even without the word "refactor."
release
Run a versioned release with git-flow discipline: deterministic preflight that reads version state from git (every tag, both branches, their remotes) rather than from the checkout, derives the semver bump from Conventional Commits, writes the changelog entry from fragments, and prints the exact command sequence for THIS repository — consent-gated by default (configurable auto mode). The workflow itself lives in .supermodo/rules/release.md, materialized once from a shipped template (light dev→main, full git-flow with release/* and hotfix/*, or request-based for a protected main). Renders git commands from verified facts and never invents a forge's — publishing is the project's own step, written into its rules file. Use for 'release', 'cut a release', 'ship a version', 'bump the version', 'publish vX.Y.Z', 'hotfix production' — whenever accumulated work should become a tagged, published version.
tdd
Test-driven work in two modes. Default: real Test-Driven Development — red-green-refactor, one failing test at a time, used while implementing features (and by the work skill). With --debug: test-first debugging — generate all plausible root-cause hypotheses, write tests to validate each before touching production code, fix based on which tests fail, then sweep the codebase for the same pattern. Use for 'tdd', 'test-driven', 'write the test first', implementing with tests leading — and with --debug whenever the user reports a bug, unexpected behavior, or failed fix attempts and the root cause is unclear.
work
Lead implementer: use the docs router to load the active task and its context chain, pick or receive the next incomplete task, implement it under project conventions, then cross-provider adversarially verify the diff. Use to resume roadmap work, start or continue a task, implement the next item, or as flow stage 2. In flow mode the interview is skipped and doc mutation is left to librarian.
bug-council
Last resort for ONE stubborn bug the ordinary attempts already failed on — a blind, evidence-driven council of Codex, Claude, Kimi, or available native subagents that falsifies competing hypotheses, implements the smallest causal fix, and verifies it independently. EXPLICIT INVOCATION ONLY - deliberately slow and token-expensive, so never auto-trigger it, never chain into it from another skill, and never run it on bugs a hunt report merely listed. Use only when the user names it (/bug-council, "convene the council", "bring in the bug council") or explicitly asks for the heaviest possible investigation of one specific bug. Otherwise, at most SUGGEST it - when a fix attempt has already failed, a bug is intermittent or unreproducible, the cause is genuinely disputed, or a regression has no obvious culprit - and wait for the user to say yes. For ordinary bugs use tdd --debug; to find unknown bugs use hunt.
next
Shows the worklist board — everything that could be worked on, grouped by priority, annotated with effort, execution state, dependencies and whether triage is still owed by the human — followed by a shortlist of three to five things worth doing now. Use whenever the user asks what to work on next, wants an overview of open work, asks to see the board, the backlog by priority, or what is blocked, what needs their input, or what is worth doing now. Also `--triage` to set priorities on items that lack one and `--repair` to report convention debt to librarian. Triggers on 'what should I work on', 'what's next', 'show the board', 'what's blocked', 'what needs me', 'prioritize the backlog'.
reports
Renders the reports supermodo has written — every flow run, every standalone skill report, every release — as self-contained HTML pages plus a four-tab archive index, and opens them in the browser. Use whenever the user asks to see a run in the browser, wants the report page, the archive, the history of what supermodo has done in this project, what is waiting on them, or asks to re-render or reopen a report. Also invoked by other skills to render the report they just wrote. Triggers on 'show me the run', 'open the report', 'the archive', 'in the browser', 'what needs my attention', 're-render the reports'.
Bio shown is the top-scored skill's repo description as a fallback — real GitHub bios land in a future update.