work

A coding workflow that carries out an existing implementation plan or fixes issues found in a code review. It divides the work into chunks and assigns them to helper agents.

In plain words
What is it for?
Use it to implement a saved plan or apply findings from a code review. It is also used to continue work from an existing Flywheel session.
Why use it?
It removes the need to manually track which planned tasks or review fixes should be handled next. It also keeps progress tied to a saved session.

Skill for Claude CodeCodex

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods
npx agentmods add skills/wsauret/flywheel/work
Any agent
npx skills add wsauret/flywheel --skill work
Clone the repo
git clone --depth 1 https://github.com/wsauret/flywheel

Made for: Claude Code, Codex.

Per session 49 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 4,351 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 1 finding. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5 $0.00049 $0.04351
Opus 5 $0.00024 $0.02176
Sonnet 5 $0.00010 $0.00870
Haiku 4.5 $0.00005 $0.00435

Measured 2d ago against content hash c44534b4fcbd, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

work scanned grade A with 1 finding against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Makes network callslowCapability

Not a fault in itself. Listed so you know the mod talks to something, and to what.

6. **Manual verification** — if `phase.manual_verification` is non-empty (plan mode), **you (the orchestrating agent) perform these checks yourself.** Do not surface them to the user. Do not call `AskUserQuestion`. You h
flywheel/skills/work/SKILL.md · 346 lines

How it starts

The opening of the file, as written. The whole thing — 346 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Work Implementation Skill

Execute the active session's plan or review findings via probe → dispatch → checkpoint. Namespace: plugin uses .flywheel/plugin/sessions/.

Input

No required arguments. Optional $ARGUMENTS:

  • empty → use the active session from .flywheel/plugin/active.json
  • slug (^[a-z0-9-]+$) → prefix-scan .flywheel/plugin/sessions/<slug>-*; tiebreak by most-recent date

Mode is detected from session contents — never set explicitly.


Phase 0: Session Detection

Active pointer lives at .flywheel/plugin/active.json ({ "schema_version": 1, "session_id": "<id>" }); sessions live at .flywheel/plugin/sessions/<session-id>/.

Decision tree:

  1. No args, active.json missing → error: "No active session. Run /plan or /work <slug>." Exit.
  2. No args, active.json points to missing session dir → error: "Session <id> not found. Clearing active pointer." rm -f .flywheel/plugin/active.json. Exit.
  3. No args, active.json presentSESSION_ID=$(jq -r .session_id .flywheel/plugin/active.json); use it.
  4. Slug arg → prefix-scan: find .flywheel/plugin/sessions -maxdepth 1 -type d -name "${SLUG}-*" | sort -r. ISO-date suffix sorts lexically = chronologically; first result wins. Update active.json atomically (.tmpmv) to the winner.

Stale-tmp cleanup: if progress.json.tmp or session.json.tmp exist from an interrupted prior write, delete them. The authoritative file is the un-suffixed one; partial writes never persist past a crash because .tmpmv is atomic on local POSIX.

Validate session.json.schema_version == 1. Mismatch → "Unsupported schema version <N>. Re-run the producing skill to regenerate."


Phase 1: Mode Detection & Load

progress.json is the file the rest of the pipeline reads to know what's happening. work-review reads it to see what files you touched. /yolo reads it to detect the fix-findings transition. Future /work invocations read it to resume. Without progress.json, the rest of the pipeline can't see your work — even if the code is correct, nothing downstream knows you ran. Treat writing it as the first thing you do, not the last.

Read the full file on GitHub · 346 lines

Files

What ships with it

1 file beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 2d ago First seen · 346 lines · 49 tokens per session scan A c44534b4fcbd

Subscribe to this mod's changes

work is a skill published in the GitHub repository wsauret/flywheel (14 stars, last pushed 5d ago), licensed MIT. It adds 49 tokens to every session and 4,351 once invoked, about $0.0002 per session on Opus 5. A static security scan graded it A with 1 finding (makes network calls). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.