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/skullninja/coco-workflow/phasegit clone --depth 1 https://github.com/skullninja/coco-workflowWhat 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 | $0.00021 | $0.01948 |
| Opus 5 | $0.00010 | $0.00974 |
| Sonnet 5 | $0.00004 | $0.00390 |
| Haiku 4.5 | $0.00002 | $0.00195 |
Grade A, and why
phase 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 — 233 lines — stays where its author put it; the contents beside it link to each section on GitHub.
User Input
$ARGUMENTS
$ARGUMENTS is a phase identifier or description. If empty, ask the user which phase to start.
Setup
- Read
.coco/config.yamlfor project configuration.
Execution
This command runs to completion in one turn. After the plan gate in step 3,
there are no further checkpoints. Do not end a turn on a statement about the next
feature ("Next I'll design X", "Continuing with feature 2") -- take the step
instead. Do not offer to pause between features. The user approved the phase; the
only legitimate stops are a hard failure or Phase Completion. Same four rules as
/coco:loop -- see The Loop Does Not Stop To Report in commands/loop.md.
Report progress as one line per feature as it resolves, so the user can see phase position without asking:
[phase "{name}" | feature {i}/{n} | {done} complete] {feature} -> {stage}
1. Identify Specs for the Phase
Read discovery.roadmap_dir from config (default: docs/roadmap).
Structured roadmap lookup (preferred):
- Glob
{roadmap_dir}/*.mdfor roadmap files - If
$ARGUMENTSmatches a phase name or number in a roadmap file (e.g., "Phase 1: Foundation"):- Parse the phase table to get the feature list with slugs, priority order, and scores
- Extract cross-feature dependencies from the phase section
- If
$ARGUMENTSis a release name (e.g., "v1.0"):- Open
{roadmap_dir}/{release}.mdand ask the user which phase to execute
- Open
Fallback (no roadmap files found):
- Search project README or docs for a roadmap section
- Use
$ARGUMENTSas a description of features to implement - Check existing spec directories in
{specs_dir}/
Extract all features for the phase. When sourced from a roadmap, preserve the priority order and scores for reporting.
2. Audit Existing State
For each feature:
a. Check for existing spec directories using the Glob tool for each feature: {specs_dir}/{feature}/*
b. Check for existing code using Grep and Glob tools to search the codebase for related files (do NOT delegate to subagents for this).
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 · 233 lines · 21 tokens per session scan A 95ada2dfae0d
phase is a command published in the GitHub repository skullninja/coco-workflow (7 stars, last pushed 2d ago), licensed MIT. It adds 21 tokens to every session and 1,948 once invoked, about $0.0001 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 commands, from other repositories
super-dev
Command "super-dev" from shangyankeji/super-dev, covering /super-dev (claude-code), 输入, super dev runtime contract, local knowledge contract and 首轮响应契约(首次触发必须执行).
validate
You are the quality gate. Your job is to verify that a phase's implementation meets quality standards before the team moves to the next phase. No phase should advance until validation passes.
implement
You are the phase dispatcher. Your job is to route implementation work to the correct phase agent with the right context.
progress
You are generating a visual progress report showing which phases are complete, in-progress, or pending for the current feature.
commit
You are creating a conventional commit scoped to the current phase of work.
prd
You are generating a Product Requirements Document (PRD) for the project described below.