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 DiUS/agent-toolkit --skill prd-to-steel-threadgit clone --depth 1 https://github.com/DiUS/agent-toolkitWrote 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/dius/agent-toolkit/prd-to-steel-thread)<a href="https://agentmods.dev/skills/dius/agent-toolkit/prd-to-steel-thread"><img src="https://agentmods.dev/badge/skills/dius/agent-toolkit/prd-to-steel-thread/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/dius/agent-toolkit/prd-to-steel-thread"><img src="https://agentmods.dev/badge/skills/dius/agent-toolkit/prd-to-steel-thread.svg" alt="Reviewed on agentmods" width="80" height="20"></a>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 | $0.00115 | $0.02909 |
| Opus 5.5 | $0.00046 | $0.01164 |
| Sonnet 5.5 | $0.00023 | $0.00582 |
| Haiku 4.5 | $0.00012 | $0.00291 |
Grade A, and why
prd-to-steel-thread 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 21d 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 — 265 lines — stays where its author put it; the contents beside it link to each section on GitHub.
User Input
$ARGUMENTS
Use the input to locate the source PRD and any existing task list. If a path is missing or ambiguous, ask for it rather than guessing.
Purpose
Convert a product-focused PRD into steel-thread.md, a delivery input for technical planning
and design. The document starts with the smallest real end-to-end user outcome that proves the
architecture works, then sequences the remaining scope as lean, demo-ready vertical slices.
steel-thread.md sits between product specification and technical planning:
PRD -> steel-thread.md -> SDD plan/design -> implementation tasks
For Spec Kit, it is an input to /speckit.plan after the feature specification exists. For
another SDD tool, hand it to the equivalent planning or design step. Do not turn the PRD into a
technical design inside this skill.
Read references/vertical-slice-method.md before slicing. Use templates/steel-thread.md for the output. If the source PRD does not follow a known structure, use templates/prd-input-template.md as an interpretation guide, not as permission to fill gaps.
Non-negotiable rules
- The PRD is the product source of truth. Do not invent requirements or silently resolve gaps.
- Ask clarification questions one at a time. Do not assume architecture, infrastructure, team capacity, delivery assignments, framework commands, or task boundaries.
- Preserve original requirement and task descriptions verbatim when mapping them to slices.
- Slice 0 is the steel thread: solo, first, deployable, and end-to-end.
- Every slice ends with a functional result that an engineer can review and a product stakeholder can demo.
- Add infrastructure only in the first slice that needs it. Never create an infra-first phase.
- Decompose only when doing so accelerates feedback, reduces risk, or makes a large story deliverable within a few days.
- Parallelise only independent work. Capacity is not a reason to force unsafe concurrency.
- Record capacity as the number of Dev+agent pairs, not personal names.
- Keep the roadmap flat as
Slice 0..N; express concurrency through dependencies, parallel-safe annotations, pair counts, and parallel groups. - Save progress in the roadmap itself, not only in conversation. An incomplete draft is not approved input to technical planning.
What ships with it
3 files 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.
- 21d ago First seen · 265 lines · 115 tokens per session scan A ab16b0ca3096
prd-to-steel-thread is a skill published in the GitHub repository DiUS/agent-toolkit (2 stars, last pushed 21d ago), licensed MIT. It adds 115 tokens to every session and 2,909 once invoked, about $0.0005 per session on Opus 5.5. 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 2026-10-09.
Other skills, from other repositories
task-creation
Write and review work items in an issue tracker: establish which reader the item is for, verify every claim it makes against the system, and give the implementer a location and acceptance criteria. Sizing and decomposition belong to tasks.
youtrack
Work YouTrack as an agent: what a project configures rather than fixes, what a write changes besides the field it names, and which failures report success. Not what a work item should say.
tasks
Work item decomposition — break implementation phases into sized, dependency-mapped tasks, each producing a verifiable artifact. Invoke whenever task involves the T stage of the DRAFT pipeline, decomposing phases into tasks, or creating a task breakdown.
handoff
Triage this conversation into a standalone transfer document — decisions, constraints, resource identifiers, verification state, and remaining work — for a session restart or a teammate.
backlog
Manage the project's tech debt backlog. Add, list, fix, wontfix, delete, prioritize, and suggest batch actions on tracked issues, run by the user to manage accumulated debt (skills file their findings through /.zuvo/backlog-add themselves, never through this skill). Modes: list [category], add [description], fix B- …
retro
Engineering retrospective from git metrics. Reports deployment frequency, release cycle span, churn hotspots, backlog health. Outputs narrative report with 3+ actionable items. Flags: --since, --path, explicit range argument.