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 skills add SteveVitali/agent-skills --skill decompose-specgit clone --depth 1 https://github.com/SteveVitali/agent-skillsWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/stevevitali/agent-skills/decompose-spec)<a href="https://agentmods.dev/skills/stevevitali/agent-skills/decompose-spec"><img src="https://agentmods.dev/badge/skills/stevevitali/agent-skills/decompose-spec/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/stevevitali/agent-skills/decompose-spec"><img src="https://agentmods.dev/badge/skills/stevevitali/agent-skills/decompose-spec.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector pass
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.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00080 | $0.05494 |
| Opus 5 | $0.00040 | $0.02747 |
| Sonnet 5 | $0.00016 | $0.01099 |
| Haiku 4.5 | $0.00008 | $0.00549 |
Grade A, and why
decompose-spec 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 — 309 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Decompose Spec
Turn one large spec into the fewest self-contained tickets that can each be implemented in a single
fresh context — then seed the durable build ledger that orchestrate-build drives to completion.
This is the planning half of a multi-session build. It does not implement anything, create branches, or run code. Its entire job is to answer "what are the units of work, in what order, and where do we cut?" — well, because the decomposition is the ceiling on the whole build: a bad split produces tickets with hidden cross-dependencies that no downstream rigor can recover from (the multi-session analogue of "the build is only as good as its spec"). Treat this as the highest-leverage, highest-rigor step, not a formality.
When to use: a spec too large for one focused implementation run, that you intend to ship as a chain of reviewable PRs across multiple sessions. For a spec that fits one sitting, skip this and run
implement-specdirectly. The per-ticket worker in both cases isimplement-spec.
The objective (what "optimal" means here)
Minimize the number of tickets N, subject to — for every ticket:
- It fits one fresh-context
implement-specrun with rigor headroom. The ticket's working set (its spec slice + the code it must read + the tests + the self-review/gap/verify passesimplement-specruns inside it) stays well under half the worker's context window. On thesubagentdispatch target this is a hard failure boundary — a subagent cannot compact, so a ticket that overflows fails. Onheadless/manualit is softer (the worker compacts, trading gap-analysis fidelity for survival). Size to the ceiling of the declareddispatch_target. - It is one coherent, independently landable concern — a single logical thing that stands alone as one CI-green PR, traceable to spec section(s). Soft proxies (not laws): ~one focused change, on the order of a few files and a few hundred lines. A large mechanical rename can be far bigger and still trivial; LOC is a proxy for cognitive size, not the target.
- It respects the dependency order — a ticket never forks from work that hasn't landed yet.
- Its boundaries fall on the weakest-coupling seams — cut where two units share the least (few cross-references, no shared implicit decisions), so each cut creates the fewest integration seams the capstone must later reconcile. High cohesion within a ticket; low coupling across.
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.
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 Changed · +23 lines e16aa1653772
- 8d ago Changed · +42 lines · +21 tokens per session 543890922e51
- 12d ago First seen · 244 lines · 59 tokens per session scan A e7590a84cd10
decompose-spec is a skill published in the GitHub repository SteveVitali/agent-skills (12 stars, last pushed yesterday), licensed MIT. It adds 80 tokens to every session and 5,494 once invoked, about $0.0004 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
neat-freak
Knowledge and governance closeout: reconcile project docs, rule files (CLAUDE.md/AGENTS.md), authorized agent memory, and workspace residue with what the code and runtime actually do, so the next session or the next person starts from one current answer. Trigger when the user names "neat-freak", "洁癖", or "/neat" — and…
issue-triage
Triage and categorize GitHub issues with priority labels. Use when user says "triage issues", "check issues", "review open issues", or during regular maintenance of GitHub issue backlog.
feature-dev
Guide a feature implementation through a structured seven-phase workflow with deep codebase understanding, clarifying questions, parallel architecture design, and quality review. Use this skill when the user asks to build a new feature, add functionality, or wants a methodical approach to implementation rather than…
product-manager-toolkit
Use when the user needs product management workflows such as RICE prioritization, customer interview analysis, PRD templates, discovery frameworks, go-to-market strategy, feature prioritization, research synthesis, or requirements documentation.
artifact-metadata
Manage artifact metadata, versioning, ownership, and review history across the SDLC lifecycle.
mention-conventions
Display @-mention naming conventions and placement rules.