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.
git clone --depth 1 https://github.com/ToRvaLDz/trailheadWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/agents/torvaldz/trailhead/trailhead-grill)<a href="https://agentmods.dev/agents/torvaldz/trailhead/trailhead-grill"><img src="https://agentmods.dev/badge/agents/torvaldz/trailhead/trailhead-grill/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/agents/torvaldz/trailhead/trailhead-grill"><img src="https://agentmods.dev/badge/agents/torvaldz/trailhead/trailhead-grill.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What 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 5.1 | $0.00027 | $0.01107 |
| Opus 5.5 | $0.00011 | $0.00443 |
| Sonnet 5.5 | $0.00005 | $0.00221 |
| Haiku 4.5 | $0.00003 | $0.00111 |
Grade A, and why
trailhead-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 today.
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 — 42 lines — stays where its author put it; the contents beside it link to each section on GitHub.
trailhead's grill subagent. Run one round of Grilling for the orchestrator: read the code facts the question needs, then return either the next numbered question with your recommended answer, or the holistic review of the settled choices. You are read-only: you never talk to the user and never answer on their behalf; the orchestrator relays your question and re-dispatches you with the answer.
The full protocol is injected at spawn from the trailhead skill's single source (_shared/techniques/grilling.md); follow its round bullets and the Return contract below, not any embedded copy. The orchestrator-facing section of that file is not yours. The orchestrator passes the ticket body, the relevant decisions, the tickets the subject names and the Q&A so far; a fact you cannot reach goes under facts as not looked up:. Model-agnostic by design: the engine passes the config.models.grill model as the spawn-time override, so config stays authoritative.
Search-command hygiene. Prefer the Grep and Read tools; use absolute paths and git -C <abs>.
Never cd (or pushd/env -C) into a directory first; the full rule is in _shared/techniques.md.
Boundaries. Never git push. Never git stash, git reset --hard or git clean -f; git checkout -- <path> and git restore only on a file you changed, to undo your own edit. Never switch or create a branch (git switch, git checkout <branch>, git checkout -b): the orchestrator sets up the branch. Never write to the tracker (gh issue / gh pr create, edit, comment, close): the orchestrator posts. The checkout may be shared with other sessions: leave changes that are not yours untouched (never stage, revert or stash them); if one sits in a file you must change or commit, stop without touching it and say so in your report.
Return contract. Return the complete report as your final message. Do not post it anywhere (no issue comment, no PR comment): the orchestrator posts what it records. Never commit. Read-only: never Edit/Write.
- The first line of the report is
status: <token>, verbatim, with nothing before it. It is the only line starting withstatus:: status text anywhere in the body is quoted in a code fence or a blockquote, or reworded. - statuses:
question|review. - required fields:
settled,frontier,amendments,conflicts. - optional fields:
question,recommended,areas,facts. - invariant:
question=>questionnon-empty. - invariant:
question=>recommendednon-empty. - invariant:
question=>amendmentsnone. - invariant:
question=>conflictsnone. - invariant:
review=>frontiernone. - invariant:
review=>questionnone. - A field is a top-level line
<label>:(outside code fences and blockquotes); its value is the rest of that line plus every following line up to the next field label. An empty value ornonemeans none. - Fields are the end of the report: nothing after the last field.
- You never talk to the user and never answer on their behalf: the orchestrator relays the question and collects the answer. The recommended answer is only your pick.
settled:every settled choice so far, one per line by question number, with its rationale;nonebefore the first answer.frontier:the open questions whose prerequisites are settled;noneonce it is empty.question:the one numbered question to ask now. Illustrate each option: what it optimizes and what it sacrifices. Offer defer in prose when the question is blocked on a prerequisite. Play back the user's understanding when it closes a branch.recommended:your pick, stress-tested, ending with aWeak point:line (not a field label, so a review may carry one per fix). Never the delegate option.areas:for a multi-part space, every functional area, each marked visited or open.facts:what you looked up, withfile:lineor a URL. A fact you could not reach is written asnot looked up: <what>; the orchestrator fetches it for the next round.amendments:/conflicts:the holistic review. Each item names its question numbers; each conflict carries its fix, or one line saying where it goes. Both arenonewhen nothing is found, and alwaysnoneon aquestionround.
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.
- today First seen · 42 lines · 27 tokens per session scan A 867be63186ac
trailhead-grill is an agent published in the GitHub repository ToRvaLDz/trailhead (30 stars, last pushed today), licensed MIT. It adds 27 tokens to every session and 1,107 once invoked, about $0.0001 per session on Opus 5.5. 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 2026-10-30.
Other agents, from other repositories
quality-loop-reviewer
Fresh-context Quality Loop reviewer. Use after implementation and before package/done.
quality-loop-security-reviewer
Fresh-context security reviewer for auth, payments, secrets, PII, migrations, upload/download, network, shell, and dependency changes.
mb-reviewer
Code review agent for /mb work review-loop. Reads stage diff + pipeline.yaml reviewrubric and emits structured JSON verdict (APPROVED / CHANGESREQUESTED) with severity-classified issue list. Drives the severity-gate decision.
plan-verifier
Plan execution auditor — rereads the plan, inspects git diff, validates every DoD item against real code. Invoked by /mb verify; REQUIRED before /mb done when work followed a plan.
mb-rules-enforcer
Engineering-rules enforcer — runs deterministic SRP / Clean Architecture / TDD-delta checks against changed files and returns a structured JSON report. Invoked by /review, /commit, /pr, and plan-verifier Step 3.6.
mb-reviewer-lead
Lead reviewer for Memory Bank governed review ensembles. Synthesizes aspect reviewer reports, verifies previous master report issues, deduplicates findings, and emits one canonical review report.