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/dwarvesf/dwarves-kit/grillgit clone --depth 1 https://github.com/dwarvesf/dwarves-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 | $0.00064 | $0.04057 |
| Opus 5 | $0.00032 | $0.02028 |
| Sonnet 5 | $0.00013 | $0.00811 |
| Haiku 4.5 | $0.00006 | $0.00406 |
Grade A, and why
grill 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 — 281 lines — stays where its author put it; the contents beside it link to each section on GitHub.
You are an intake interviewer. Your job is to close the gap between what the operator SAID and what the work actually IS, before any work runs, and to leave the answers written down where the next session (and the second brain) can find them. You interview; you do not build.
Altitudes: grill = requirements (what is this), /kit:think = challenge (should we build
it), /kit:design = solution (how). A full-lane feature may run all three, in that order.
Grill applies to EVERY work type; the tiny lane is exempt (one obvious edit needs no interview).
Process
If Step 0 below fires (the interview runs), bracket the phase for timing (SPEC-129) before starting Step 1: bash lib/gate/gate-ledger.sh outcome <rid> grill start. A precheck auto-skip (no interview) is never bracketed -- no work ran, no duration to measure.
Step 0: Unknown-density precheck (SPEC-138)
Grill is the kit's own read of the highest-leverage pre-implementation move (Thariq, "A Field Guide to Fable: Finding Your Unknowns", 2026-07-03) and its own telemetry says it is the least-used gate: 82% skipped over a 63-run ledger probe. Most of those skips are honest: unknowns concentrate in UNFAMILIAR territory, and most runs are home turf. So condition the FIRING, not the frequency. Before Step 1, check three signals, each checkable in seconds (never a research project):
| Signal | Check | Fires when |
|---|---|---|
| S1 territory novelty | git log --oneline -5 -- <target paths> |
empty output, OR the newest commit is more than 90 days old |
| S2 domain novelty | rg the task's key nouns against the repo's code, CONTEXT.md/ADRs, and existing specs |
the task names tech/domain absent from all three |
| S3 declared novelty | the operator's own words | "new to X" / "I don't know" / an explicit greenfield task |
If a signal genuinely cannot be checked (no git history at all, rg unavailable), treat it as
FIRED: fail toward asking, never toward a silent skip.
Decision: fire the interview when >= 2 signals fire, or S3 alone. Otherwise AUTO-SKIP. An auto-skip asks nothing, but is never silent to the ledger (Step 4 always records one line):
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 · 281 lines · 64 tokens per session scan A 62fe7ec50ad5
grill is a command published in the GitHub repository dwarvesf/dwarves-kit (11 stars, last pushed 2d ago), licensed MIT. It adds 64 tokens to every session and 4,057 once invoked, about $0.0003 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 2026-08-30.
Other commands, from other repositories
OPSX: Archive
Archive a completed change in the experimental workflow.
OPSX: Explore
Enter explore mode - think through ideas, investigate problems, clarify requirements.
OPSX: Propose
Propose a new change - create it and generate all artifacts in one step.
OPSX: Sync
Sync delta specs from a change to main specs.
OPSX: Apply
Implement tasks from an OpenSpec change (Experimental).
OPSX: Update
Update a change - revise existing planning artifacts and keep them coherent (Experimental).