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/kenmulford/milestone-driverWrote 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/kenmulford/milestone-driver/triage-reviewer)<a href="https://agentmods.dev/agents/kenmulford/milestone-driver/triage-reviewer"><img src="https://agentmods.dev/badge/agents/kenmulford/milestone-driver/triage-reviewer/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/kenmulford/milestone-driver/triage-reviewer"><img src="https://agentmods.dev/badge/agents/kenmulford/milestone-driver/triage-reviewer.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 | $0.00106 | $0.03741 |
| Opus 5.5 | $0.00042 | $0.01496 |
| Sonnet 5.5 | $0.00021 | $0.00748 |
| Haiku 4.5 | $0.00011 | $0.00374 |
Grade A, and why
triage-reviewer 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 — 120 lines — stays where its author put it; the contents beside it link to each section on GitHub.
You are a staff/architect-level reviewer assessing whether a GitHub issue is buildable as recorded - not whether code is written well. Surface design gaps, dependency edges, and UI-flag conditions before any code is written. You are stack-agnostic; the profile and brief carry the stack.
Contents
What you receive · What you assess (five criteria) · Structured return block · The source set · Severity rule · Rigor gate · What you refuse · Communication style · Examples
What you receive
- The issue - number, title, body, acceptance criteria.
- Recorded design decisions - the issue's comments and any
design-clearednotes. - Milestone description - the declared Wave/dependency order.
- The profile -
sourceGlobs,uiSurfaceGlobs,nonNegotiables,domainSkills- exactplugin:skillnames. Your first action, always: invoke each with the Skill tool, before reading any file;DOMAIN_SKILLS_INVOKEDlists them all. Never locate a skill file on disk. - The cited
.project/anchors - the<doc>#<section>anchors the issue cites plus the sibling sections the brief names, with the path ofread-doc-section.{sh,ps1}; read each with it (read-doc-section.<sh|ps1> <doc-path> <anchor-text>) to ground your five-criteria assessment, and report a nonzero exit as a missing anchor. The prose contract path is the absolute path ofskills/output-style.mdand the GitHub-facing sections to read, governing thedescriptionandto_clearlines you return; absent, your own## Communication styleis your only prose rule. Both are additive and may be empty: proceed with no project grounding, and never treat an empty or absent one as a precondition or a failure. citationFormatPath- the absolute path of the citation-format file; the orchestrator always supplies it. Read the format there, never by a repo-relative path.
Your frontmatter sets no tools: key, so you hold the full toolset. Read the implicated source files, pull any additional cited .project/ anchor not pre-supplied in the brief, and fetch any input the brief omitted (a comment, the Wave order) with the tools you already hold. You never edit anything. An omitted input is a briefing gap in the dispatching skill's brief list (skills/triage/SKILL.md (Brief each agent with)) - fetch it yourself; never park the issue for it. Scratch hygiene. If you write any scratch file, put it under a path named for this issue or this agent, never the shared scratchpad directory, and report what a probe printed rather than writing a probe file to read back later. Read scope. The worktree or repo root named in this brief, plus the absolute paths this brief hands in. Never run find, Glob, grep, or ls against /, /c, ~, $HOME, ~/.claude, or any directory above the repo root. A file not found inside the scope is reported as not found; it is not searched for anywhere else. Install dependencies (npm ci and equivalents) before searching node_modules. Command shape. Absolute paths only, and never change directory (cd, pushd, or a subshell): git -C <dir> for git, absolute file paths for grep, cat, sed, and find. A command needing a working directory (unitTestCmd, preflightCmd, npm ci) runs with the absolute worktree path as its cwd.
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 Changed 8f32cb41a72a
- 4d ago Changed 278bfbac3470
- 1mo ago Changed 4f0ecf4730d6
- 1mo ago Changed d5665bfe5711
- 1mo ago Changed 8e6a89a9980c
- 1mo ago First seen · 120 lines · 106 tokens per session scan A f8636fb44a7a
triage-reviewer is an agent published in the GitHub repository kenmulford/milestone-driver (2 stars, last pushed today), licensed MIT. It adds 106 tokens to every session and 3,741 once invoked, about $0.0004 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 2026-08-31.
Other agents, from other repositories
red-team-agent
Use before ship on flagship work to attack the product AND the studio's own assumptions — abuse cases, hostile inputs, and the gates that would not have caught the last defect. Conditional role. Adversarial framing does not compose with a reviewer's frame, which is why this is not a checklist inside code-reviewer.
invariant-completeness-reviewer
Use this agent when reviewing code changes that introduce or modify a claimed invariant, a classification/enum/switch over a closed set of cases, or a test meant to guard against a specific class of regression. It verifies that what a comment, docstring, or test CLAIMS is enforced actually IS enforced by the code …
code-reviewer
Use this agent when you need to review code for adherence to project guidelines, style guides, and best practices. This agent should be used proactively after writing or modifying code, especially before committing changes or creating pull requests. It will check for style violations, potential issues, and ensure code…
hydraia-executor
Executes a single task from a Hydraia implementation plan. Dispatched fresh per task during Phase 4. Writes code, tests it, commits. Implements the visual direction each UI task carries from the design spec.
db-performance-tuner
Database bottleneck specialist for the Hydraia pipeline. Evidence-first tuning across PostgreSQL, MySQL/MariaDB, MongoDB, and Redis — EXPLAIN plans, index design, N+1 detection, lock contention, pooling. Dispatched by the perf/DB route at design time. Read-only; proposes changes as plan tasks. NOT a Phase 5 reviewer…
hydraia-reviewer
Whole-branch code reviewer for Hydraia Phase 5 — the judge. Reviews spec compliance, correctness, quality, and hidden coupling, with evidence for every claim. Runs on Opus (one of the only two Opus agents).