Getting it into your agent
This one installs as part of its plugin. Adding the marketplace and installing the plugin brings it with everything else the plugin ships.
/plugin marketplace add ihudak/ihudak-claude-plugins/plugin install dev-workflowsWrote 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/commands/ihudak/ihudak-claude-plugins/implement)<a href="https://agentmods.dev/commands/ihudak/ihudak-claude-plugins/implement"><img src="https://agentmods.dev/badge/commands/ihudak/ihudak-claude-plugins/implement/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/commands/ihudak/ihudak-claude-plugins/implement"><img src="https://agentmods.dev/badge/commands/ihudak/ihudak-claude-plugins/implement.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 | $0.00048 | $0.28326 |
| Opus 5.5 | $0.00019 | $0.11330 |
| Sonnet 5.5 | $0.00010 | $0.05665 |
| Haiku 4.5 | $0.00005 | $0.02833 |
Grade A, and why
implement 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 23d 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 — 999 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Implement the following: $ARGUMENTS
Core references. A citation of the form workflows-core:<name> names a shared reference in the workflows-core plugin. Load it with Skill(skill: "workflows-core:reference", args: "<name>") — never by path: ${CLAUDE_PLUGIN_ROOT} resolves to this plugin, which does not carry it.
Phase 0 — Load and classify inputs
Strip --no-commit first, before any other parsing: it is the only flag this command takes, and an unstripped flag is read as free-text prose and lands in the task description. When present, Phase 4.6 is skipped entirely and the changes are left in the working tree. It is the one opt-out from a commit that is otherwise prompt-free (${CLAUDE_PLUGIN_ROOT}/references/code-handoff.md §1 rule 5) — typed rather than clicked, which is the difference — and a run under it says once what it costs: the work is recoverable only on this machine, and /document and /release-notes will not find it later.
$ARGUMENTS may contain free-text prose plus zero or more @path tokens (today's single-@file form is a subset). Resolve each @path relative to the current working directory. Classify each @path — and the current working directory — by inspection, not by matching the path string:
| Detected as | Recognition rule | Handling |
|---|---|---|
| Spec file | a single .md file |
read fully; use as the description/spec |
| Spec folder | a directory containing prompt.md and/or a *-design.md |
read all .md specs within; fold into the description |
| Specs folder | a directory under specifications/ that resolve-address resolves — its kind: and key: read off the folder's carrier (workflows-core:addressing §4) |
hand to the folder read in Phase 1.7 |
| Code repo | a directory where git -C <path> rev-parse --is-inside-work-tree succeeds (includes the cwd) |
scan target in Phase 1.7 |
Address resolution. Before the per-@path classification above, look for a single positional
address in $ARGUMENTS — a <KEY>, or an @<path> naming a folder in the specs tree. Present →
resolve it with resolve-address (Skill(skill: "workflows-core:reference", args: "addressing resolve-address"), §3) and the run
is keyed; absent → the run is direct (free-text / @file, this command's existing flow).
That is the whole mode test, and it unifies the input grammar with /document. On a <KEY>,
$SPECS_PATH comes first: if it is unset, stop naming it before resolving anything
(choices: ["Set SPECS_PATH (enter the path)", "Cancel"], workflows-core:escalation-rules
Required path environment variable unset) — a key is found only by searching the specs tree, so
with no tree the absent stop below would name the wrong cause and offer a re-enter that cannot
succeed. An @<path> address needs no specs tree: it is resolved without the variable and runs on,
its record written into the folder it names, and a direct run is unaffected. ambiguous is §3's
hard stop, naming every match. status: absent is a stop, never a fall-through to direct mode
— the key dir not found rule in workflows-core:escalation-rules (["Re-enter key", "Cancel"]),
naming what creates a folder this command reads: a PRD- folder comes from
/product-workflows:idea <KEY> or /product-workflows:create-prd <KEY> on the idea route and from
/product-workflows:brd-split on its parent BRD on the BRD route; an EPIC- folder from
/product-workflows:epics <PRD-ADDRESS>. A keyed run with no folder has nowhere for Phase 4.7's
record to go.
The classification table above still applies to every other @path token: a spec folder contributes
to specs, a code repo is an /implement-only scan target. Carry mode (keyed | direct), the
resolved path, kind and key, and specs forward.
Epic-unit resolution (keyed runs). /implement implements one Epic at a time. When
mode: keyed, place the resolved folder as workflows-core:addressing §4.1 does — by its prefix,
never by the kind it asserts, which on a BRD-route slice is brd — taking its container test first.
A BRD container — a BRD- folder, or a folder with no prefix holding coverage-ledger.md or
brd/brd-inventory.md and no brd-link.md naming a parent: — holds no Epic to implement, because
a BRD's work is authored in its slices; stop, before anything is read:
IMPLEMENT_BRD_NOT_SLICED: <KEY> resolves to a BRD container at <path> — a BRD's Epics belong to its PRD- slices. <the remedy>
<the remedy> lists the slices under it, found by the positive test §4.1 names — Implement within a slice instead: '/dev-workflows:implement <SLICE-KEY>' — <each slice's key>. — and, where it finds none:
It has no slice yet: '/product-workflows:brd-split <KEY> "<how to cut it>"' carves one — the instruction is required there, and that run carves nothing where this BRD's ledger leaves no row unallocated. Where it leaves none, coverage-ledger-format.md §5 names two repairs, the narrow one first: hand-edit the one row to be built back to unallocated in coverage-ledger.md, leaving every other row as it stands; or, to re-take the whole inventory, re-run '/product-workflows:brd-intake <KEY> @<brd-file>', which reopens every row wherever its read finds a requirement and discards every deferred-to, rejected and superseded-by the ledger records.
It is a user halt. A folder with no prefix — one §5's legacy fallback resolved, or an unprefixed
folder an @<path> names, a name beginning with a kind token being prefixed only where it begins
<KIND>-<the resolved key>- — is otherwise placed by positive evidence: a resolved kind: epic
counts as an EPIC- folder below, and a resolved kind: prd, or a brd-link.md naming a
parent:, as a PRD- folder. A folder none of these places is not guessed at — stop, naming the
folder and what it carries — and, where it holds an idea.md and no prd.md, name /product-workflows:create-prd <KEY> too, whose prd.md places it (workflows-core:addressing §4.1). Then:
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.
- 23d ago Changed · +15 lines cb6d25bcccf4
- 24d ago Changed · +60 lines ef33fe16da02
- 28d ago Changed · +52 lines 655c73829d95
- 1mo ago Changed · +1 lines 1511b5e362da
- 1mo ago Changed · -8 lines 62bc47bf43fa
- 1mo ago First seen · 879 lines · 48 tokens per session scan A 39b12a6ef60d
implement is a command published in the GitHub repository ihudak/ihudak-claude-plugins (2 stars, last pushed 24d ago), licensed MIT. It adds 48 tokens to every session and 28,326 once invoked, about $0.0002 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 2026-09-05.
Other commands, from other repositories
check-fix
Execute quality gates, fix any issues found, and create a single high-quality Conventional Commit summarizing all changes made.
gate-check
List or run the verify commands this repo requires for the current changes.
validate
Command "validate" from Vimalk0703/shipworthy, covering /validate — pre-push validation gate, what this does, steps, when to run and output format.
implement-feature
Implement an approved feature plan with fresh-context slices, TDD, revision-bound evidence, and PR-ready output.
verify-setup
One-time setup for /verify and /break: sniff boot/seed/health, confirm, write .verify/setup.json and the app profile.
autotask
Execute development task autonomously from description to PR-ready - handles implementation, testing, and git workflow without supervision.