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/chappygo-os/atomic-spec/_subagent-discoverygit clone --depth 1 https://github.com/Chappygo-OS/Atomic-SpecWhat 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 | $0.00000 | $0.00992 |
| Opus 5 | $0.00000 | $0.00496 |
| Sonnet 5 | $0.00000 | $0.00198 |
| Haiku 4.5 | $0.00000 | $0.00099 |
Grade A, and why
_subagent-discovery 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 3d 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 — 73 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Dynamic Subagent Discovery Protocol
Shared sub-routine — referenced by
/atomicspec.plan,/atomicspec.tasks, and/atomicspec.implement.Constitutional basis: Per the framework's design (see
CLAUDE.md), subagents are matched to tasks by semantic similarity between task keywords and each agent'sdescriptionfrontmatter. Hardcodedfolder → agentrouting tables are FORBIDDEN because they break when a consumer project adds new subagents, renames existing ones, or targets a platform (e.g., mobile) whose agent set differs from the web defaults.
When to run this protocol
Whenever a command needs to pick a subagent for a task (planning, task generation, or implementation execution). Do not hardcode agent names in command templates. Do not assume a specific agent exists.
The protocol
Step 1 — Scan
List every **/*.md file under .specify/subagents/ recursively. Exclude any file whose name starts with _ (those are shared templates or routing hints, not agents).
Step 2 — Extract
For every discovered file, read the YAML frontmatter and capture:
name— the subagent identifierdescription— what it does and when to use itmodel(optional) — preferred model
Build an in-memory list: [{name, description, path}].
Step 3 — Derive task keywords
From the task under consideration, extract:
- Objective verbs and nouns from the task title and description (e.g., "create repository", "add API endpoint", "wire React component")
- File paths in "Files to Create / Modify" (e.g.,
repositories/,routes/,components/,migrations/,workers/) - Technical terms in the implementation steps (e.g., "SQL", "WebSocket", "OAuth", "Stripe", "push notification", "Core Data")
- Platform signals from
plan.md(e.g., iOS / Android / web / backend) — these narrow the candidate pool
Step 4 — Match and score
For each candidate agent, score by keyword overlap between the task keywords and the agent description:
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.
- 3d ago First seen · 73 lines · 0 tokens per session scan A ea73e033f551
_subagent-discovery is a command published in the GitHub repository Chappygo-OS/Atomic-Spec (10 stars, last pushed 10d ago), licensed MIT. It costs nothing until one of its globs matches a file; then it loads 992 tokens. 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
taskstoissues
基于现有设计制品，将任务转换为可执行、带依赖顺序的 GitHub Issues。.
cross-critique
Adversarial multi-angle critique. All three auditors fire in parallel (codex=technical, gemini=strategic, claude=ux). Counter-args ranked by rebuttal survival score, not raw severity. Usage: /cross-critique [--with | list | compare].
ijfw-audit
Run the IJFW audit gate for the current workflow phase. Usage: /ijfw-audit [phase name].
ijfw
IJFW command index. Groups commands by intent: Build / Remember / Ship / Review / Configure.
ijfw-execute
Jump directly to the IJFW workflow Execute phase (Deep D4 / Quick Q3). Usage: /ijfw-execute [task or phase name].
full-review-gate
Perform a comprehensive repository code review in an isolated worktree. Fix P0/P1 findings and ship them as a PR targeting the current branch.