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 commands/dsnger/dev-workflow-kit/workflow-initgit clone --depth 1 https://github.com/dsnger/dev-workflow-kitWhat 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 | $0.00022 | $0.21133 |
| Opus 5 | $0.00011 | $0.10566 |
| Sonnet 5 | $0.00004 | $0.04227 |
| Haiku 4.5 | $0.00002 | $0.02113 |
Grade B, and why
workflow-init scanned grade B with 1 finding 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 yesterday.
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.
Reads agent configuration directoriesmediumAgent snooping
.claude/, .codex/, .gemini/ hold keys, settings and other credentials a mod has no legitimate need for.
Codex reads `~/.codex/config.toml` — the user's home directory, not this repo. Do not How it starts
The opening of the file, as written. The whole thing — 1,465 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Scaffold this project's cross-model review workflow: write the per-project files
below, then interactively author AGENTS.md, then print the stack-specific
checklist that stays the user's job.
Target model: Claude via Claude Code. This command is a prompt artifact and follows
the checklist it scaffolds (docs/prompt-standards.md).
Rules
- Idempotent. Run it twice and the second run changes nothing it already wrote.
- Never overwrite without asking. For each target: missing → write it. Present
and byte-identical to the template → report
unchanged, touch nothing. Present and different → show what differs (a short diff, not the whole file) and ask: overwrite / merge / skip. This matters because these files accumulate real project content after the first run — a silent overwrite destroys it. - Additive files are merged, never rewritten.
.gitattributes,.mcp.json,pnpm-workspace.yamlandpackage.jsonbelong to the project; add the missing line or key and leave everything else exactly as it was. - Stack-specific content is marked, not guessed. Where a template has a
# TODO(stack):marker, leave the marker in and name it in the closing checklist. A plausible-looking command that was never run is worse than an honest TODO. - Report what happened per file —
written/unchanged/merged/skipped (user)/asked, overwrote. No silent no-ops.
Step 1 — Preflight: check every prerequisite, then say what's missing
This is the project's one setup entry point, so it carries the whole "is my setup complete?" answer. Check each item below, state the result for every one (the detect-and-state rule governs this step — a silent pass is indistinguishable from an unrun check), and carry every failure into the closing checklist as a named blocker.
Check, in order:
- git repository —
git rev-parse --show-toplevel. If absent, stop and ask: the hook's.context/state directory and the.gitattributesunion merge both assume one, so nothing below is meaningful without it. - superpowers plugin — look for
superpowers:brainstorming,superpowers:writing-plansandsuperpowers:executing-plansin the skills available to you right now. This is a hard prerequisite, not a nicety: the workflow's whole middle (spec → plan → execute) is those skills. Without themintakehands off to a skill that does not exist, and the gate hook's Gate-A pass counter — which only resets on those skill events — silently never fires, so Gate A degrades to no enforcement at all while still looking enforced. If they are absent, reportMISSINGwith the install commands:claude plugin marketplace add obra/superpowers-marketplacethenclaude plugin install superpowers@superpowers-marketplace. - Codex MCP — the gates call
mcp__codex__exec(Gate A, reviews text) andmcp__codex__review(Gate B, reviews a diff). These three states have different causes and different fixes, so report the one that actually holds — a single "codex: failed" sends the user hunting in the wrong place:
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.
- yesterday First seen · 1,465 lines · 22 tokens per session scan B 4869809943c2
workflow-init is a command published in the GitHub repository dsnger/dev-workflow-kit (5 stars, last pushed 6d ago), licensed MIT. It adds 22 tokens to every session and 21,133 once invoked, about $0.0001 per session on Opus 5. A static security scan graded it B with 1 finding (reads agent configuration directories). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other commands, from other repositories
brooks-audit
Run a Brooks-Lint architecture audit.
simplify
The over-engineering review: five tags (delete, stdlib, native, yagni, shrink), a mandatory replacement per finding, and a real null result when there is nothing to cut.
git
The pre-finish status: branch, hygiene findings, message checks, workflow lint, template state.
init
Install the formatters this repository needs, with every command visible before it runs.
VibeGuard: ExecPlan
Long-term task execution plan — generates self-contained execution documents from SPEC, supports cross-session recovery.
VibeGuard: Review
Structured code review - first run the guard to obtain the baseline, then review according to security → logic → quality → performance priority.