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.
npx agentmods add skills/tercel/spec-forge/decomposenpx skills add tercel/spec-forge --skill decomposegit clone --depth 1 https://github.com/tercel/spec-forgeWhat 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.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00059 | $0.02263 |
| Opus 5 | $0.00030 | $0.01131 |
| Sonnet 5 | $0.00012 | $0.00453 |
| Haiku 4.5 | $0.00006 | $0.00226 |
Grade A, and why
decompose scanned grade A with 0 findings 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.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 213 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Decompose — Project Scope Analysis
Analyze project scope and determine whether to treat it as a single feature or split into multiple sub-features before entering the spec chain.
Core Principles
- Lightweight: a brief interview, not a deep investigation — ask only what you cannot already infer
- Boundary-focused: Only care about scope boundaries and dependencies
- No demand validation: That is /idea's responsibility
- No deep research: That is /prd's responsibility
- Single is valid: Not every project needs splitting — a single-feature verdict is a good outcome
Workflow
Step 1: Read Context
Scan the project to understand what exists:
@../shared/project-context.md
Execute PC.1 (Project Discovery) and PC.3 (Project Profile). Prefer the fast path — resolve <sf_scripts> (see @../shared/scripts.md) and run python3 "<sf_scripts>/sf-scan.py" --root "<project_root>", which returns the project structure, languages, frameworks, DB/auth signals, and the existing-doc inventory in one pass. Fall back to the manual PC.1/PC.3 scan only if python3 is unavailable or the script is not found — switch to the fallback silently.
- Project structure, README, existing docs (PC.1) — read from the scan's document inventory and source-tree fields
- Project profile — Web API, CLI, Frontend, etc. (PC.3) — the scan gives the raw signals; you label the profile
- Check if
ideas/{feature-name}/draft.mdexists — if found, read it for context on scope
The project profile informs the decomposition: a monorepo with multiple apps has different split criteria than a single-purpose API. Summarize what you learned in 2-3 sentences. Do not present this to the user — it is internal context for the interview.
Step 2: Scope Interview
Use AskUserQuestion to understand project boundaries. Ask only what you cannot already infer from the scan and the user's description, adapting based on answers — there is no fixed number of rounds.
Round 1 — The Shape:
- What are the main functional areas or modules of this project? (e.g., "auth, payments, notifications" or "it's a single API endpoint")
- How big is this project in your mind — a single focused feature, or multiple related features?
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.
- 2d ago First seen · 213 lines · 59 tokens per session scan A 3a775aa14283
decompose is a skill published in the GitHub repository tercel/spec-forge (5 stars, last pushed 1mo ago), licensed MIT. It adds 59 tokens to every session and 2,263 once invoked, about $0.0003 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other skills, from other repositories
ppt-design-skill
Design, generate, review, and revise editable PowerPoint presentations through a rigorous brief-to-PNG workflow using the public pptx-designer Python library.
manager
Sync session work into GitHub issues, or query track status across repos. Write: find/update issue + W-label. Read: "что по ". Triggers: "создай issue", "синкни сессию", "manager". Not day/week plans (daily-tasks).
weekly-retro
Use when the user is closing an ISO week, reviewing planned outcomes against evidence, interviewing work Areas, resolving unfinished commitments, or asking for "ретро", "weekly retro", "week review", or lessons from the week.
art-director
Orchestrate iterative visual style searches with branch prompts, decision graphs, feedback loops, and final direction selection.
corp-doctor
Use when a Personal Corp operating loop needs setup, repair, a new department, or task routing: HQ files and agent rules, GitHub issue workflow, corp- owner map, department repositories, or deciding which repo an issue belongs to. Triggers: "corp doctor", "почини контур", "заведи отдел", "куда положить задачу"…
html-draft
Use when user wants a standalone HTML diagram in flat engineering blueprint style — architecture diagrams, system flows, technical spec sheets, component maps. Generates one HTML file using Tailwind v4 (browser CDN) for layout and D3 v7 (CDN) for SVG diagrams. User-invoked only — do NOT auto-trigger. Triggers on…