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/ignaciomarin402/spec-flow-plugin/modelsgit clone --depth 1 https://github.com/IgnacioMarin402/spec-flow-pluginWhat 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 | $0.00031 | $0.00626 |
| Opus 5 | $0.00015 | $0.00313 |
| Sonnet 5 | $0.00006 | $0.00125 |
| Haiku 4.5 | $0.00003 | $0.00063 |
Grade A, and why
models 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.
What it actually says
Run:
node ${CLAUDE_PLUGIN_ROOT}/scripts/model-routing.mjs
Show its output to the human as it came out. Do not re-derive any of it. The
whole point of that script is that three layers decide an agent's model and no
single file shows more than one of them — the tier its frontmatter ships, the
project's override in .claude/spec-flow.config.json, and what
ANTHROPIC_DEFAULT_<TIER>_MODEL pins that tier to. Reading one and concluding
is how you get a confident wrong answer, which is the thing the report exists
to prevent. If it exits non-zero the routing block is unusable and every
spec-flow spawn is being denied; relay what it says and stop.
With arguments
$ARGUMENTS is empty for the report above. Otherwise it is <agent> <tier>
or <agent> default, and you edit .claude/spec-flow.config.json in the
consuming repo:
<agent> <tier>— setagents.<agent>to that tier. Create the file, or theagentsobject, if either is missing; leave every other key alone,max_opus_callsincluded.<agent> default— delete that agent's entry, falling back to what the plugin ships. Delete theagentsobject too if it ends up empty, rather than leaving{}behind.
Then run the script again and show the new report. That second run is the confirmation — it reads the file back through the same code the spawn hook uses, so "I wrote it" and "it applies" are the same statement rather than two.
Expect the edit to ask the human for approval: .claude/ is a protected
directory, and a write there prompts even when file edits are otherwise
accepted. That is the harness, not a misconfiguration — say so rather than
looking for a way around it, and never reach for a permission bypass to
land a preference. In a non-interactive session where no one can approve,
report the exact change you would have made and stop.
Do not guess an agent name or a tier. The report lists every valid agent, and a
tier is one of opus, sonnet, haiku, fable. A model id is never valid
here — a spawn accepts only a tier, and pinning a concrete version is Claude
Code's own ANTHROPIC_DEFAULT_<TIER>_MODEL setting, per session rather than
per agent. If the human asks for a version, say that and offer the settings
entry instead of writing something this contract will refuse.
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 · 51 lines · 31 tokens per session scan A 5e960e48dc50
models is a command published in the GitHub repository IgnacioMarin402/spec-flow-plugin (2 stars, last pushed 2d ago), licensed MIT. It adds 31 tokens to every session and 626 once invoked, about $0.0002 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other commands, from other repositories
xdo
Direct development by the Main Agent.
init
First-run teamctl onboarding — from no-teamctl-installed to a running supervised team in one conversation.
estate-blueprint
Generate the 4-layer Estate Blueprint (Persona / Topology / Kernel / Modules) for a sovereign intelligence estate / agent army commission, grounded in the client's Genius Profile + Freedom Path + Vision. Emits attested Blueprint doc + architecture-options matrix + 30/90-day targets + build brief. For Route D elevated…
lavra-work-teams
Work on multiple beads with persistent worker teammates that self-organize through a ready queue.
speckit.plan
When executing this command, delegate independent work to parallel sub-agents to reduce total execution time.
dev-task
Create an agent team to develop: Developer teammate + Code-tester teammate + QA-tester teammate + Reviewer teammate + Researcher subagent → implements .dev/cdt/plans/plan-$TIMESTAMP.md in waves.