The dataset viewer is not available because its heuristics could not detect any supported data files. You can try uploading some data files, or configuring the data files location manually.
YAML Metadata Warning:empty or missing yaml metadata in repo card
Check out the documentation for more information.
SPECWORKS — Factory-Style Spec Organization
A filesystem factory for building with agents. Specs are not documents that describe production — they are the production line. Work enters as an order, gets its own station, moves down the assembly line, passes quality control, and ships — with a paper trail at every station.
The floor
specworks/
├── AGENT.md # fleet constitution + FACTORY PROTOCOL
├── orders/ORD-<n>-<name>/ # order book: PRD + SCOPE + SPEC + EVAL + qc/ + ship/
├── stations/STATIONS.md # index of live branches/worktrees
├── tasks/pending/ · done/ # the bus: ALL job files (T-<n>-j<k>-...)
├── knowledge/ # shared truth: INDEX, _provisional, _graveyard, ...
├── workspaces/agents/<name>/ # per-agent drafts, isolated by lane
├── workspaces/_templates/ # order + task + verdict starters
├── skills/ # job playbooks: J-1..J-5, refresh-index, ...
├── scripts/ · hooks/ # deterministic ingest + event triggers
├── DECISIONS.md # append-only factory-wide decisions
├── ACTIVITY.md # append-only, attributed @agent/human
└── logs/ # per-session, per-agent summaries
Quickstart (day one)
- Read
AGENT.md— the roster, the write lanes, the factory protocol. - Run ORD-001 (
orders/ORD-001-factory-bootstrap/) — the factory's first order is standing up the factory itself. - Cut a station per the first real order via
skills/job-station.md. - Every handoff is a task file in
tasks/pending/— never a chat message. - QC verdicts land in
orders/<order>/qc/as immutable evidence. - Weekly:
skills/refresh-index.md+ the auditor pass (J-5).
Conventions
| Entity | Pattern | Where |
|---|---|---|
| Order | ORD-<n>-<name> |
orders/ |
| Station | station/<order-id>-<short> |
branch/worktree + stations/STATIONS.md |
| Job/task | T-<n>-j<k>-<from>-to-<to>-<topic> |
tasks/pending/ → done/ |
| QC verdict | QC-<n>-<verdict> |
orders/ORD-n/qc/ (immutable) |
| Decision | D-<n>, supersedes chaining |
DECISIONS.md (append-only) |
| Activity | date time — @actor — action |
ACTIVITY.md (append-only) |
| Session log | YYYY-MM-DD-<agent>-<topic>.md |
logs/ |
The traceability spine
PRD FR-3 → SPEC task T-042 → station commit → QC verdict → RELEASE-NOTES → DECISIONS
why what to build what was built how verified what shipped what we learned
The loop
QC fail → FEEDBACK task to the builder (error + log line + screenshot + task id) → fix → full suite → re-inspect → max 3 loops → ESCALATE to human with full history. Factory health metric: loops-to-ship per order — rising loops mean fix the SPECS, not the builders.
Ground rules
knowledge/is the only place truth lives; every fact cites a source.- Nothing is deleted, only moved (wrong knowledge →
_graveyard.md). - QC is a different agent than assembly. The builder never ships its own work.
- One order = one station = one J-1..J-4 chain. No shared stations.
- No order enters the floor without PRD + SCOPE + SPEC (with task map) + EVAL.
- Downloads last month
- 269