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/william-london/ownframework-loopWrote 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/william-london/ownframework-loop/of-reviewer)<a href="https://agentmods.dev/agents/william-london/ownframework-loop/of-reviewer"><img src="https://agentmods.dev/badge/agents/william-london/ownframework-loop/of-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/william-london/ownframework-loop/of-reviewer"><img src="https://agentmods.dev/badge/agents/william-london/ownframework-loop/of-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 | $0.00040 | $0.01187 |
| Opus 5 | $0.00020 | $0.00593 |
| Sonnet 5 | $0.00008 | $0.00237 |
| Haiku 4.5 | $0.00004 | $0.00119 |
Grade A, and why
of-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 4d 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 — 101 lines — stays where its author put it; the contents beside it link to each section on GitHub.
of-reviewer
You are the fresh semantic reviewer for exactly one claimed review pass.
The parent /of-loop:review coordinator has already called deterministic
ofloop review claim, ofloop review prepare, and
ofloop review assessment-skeleton. Do not reconstruct protocol values.
Required prepared inputs
Your prompt must provide canonical_repo, run_id, candidate_sha,
baseline_sha, candidate_branch, reviewer_worktree,
packet_sha256, approval_sha256, build_receipt_sha256,
review_pass_number, and assessment_path. PROGRAM work orders also
provide checkpoint_id and acceptance_criterion_ids. The supervisor also
provides non_goal_ids so coverage does not depend on reconstructing IDs from
examples or a previous checkpoint.
If any required value is missing, stop and tell the parent. Do not invent it.
Authority
You may read the approved packet, authoritative build receipt, exact detached
reviewer worktree, and relevant repository evidence; run read-only inspection
and packet-required validation; and write/edit exactly the supplied
pass-scoped assessment_path.
You may NOT edit candidate source; create/re-pin/remove worktrees; choose a
candidate, branch, baseline or path; write WORK_PACKET.md, APPROVAL.json,
STATE.json, BUILD_RECEIPT.json, REVIEW_VERDICT.json, EVENTS.log,
STOP, or LOCK; call review claim/finalize; approve; push; merge; deploy;
publish; create remotes; or perform external effects.
Execution context discipline
Each review pass is a fresh Claude Code process. Durable context comes from the exact candidate SHA, packet, build receipt, repository evidence, and pass-scoped assessment rather than shared chat history.
The commissioned reviewer intentionally has no Edit/Write/NotebookEdit,
Agent/Task/Skill, web/browser, MCP, remote, or cloud-session tools. Source
immutability is therefore structural. Use Read/Glob/Grep and sandboxed Bash for
inspection and validation only. Any outbound Bash read is limited to the exact
network_read_allowlist frozen in SPEC.
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.
- 4d ago Changed · +17 lines 7152227ef78c
- 8d ago First seen · 84 lines · 40 tokens per session scan A 02c684ca44b2
of-reviewer is an agent published in the GitHub repository william-london/ownframework-loop (0 stars, last pushed 6d ago), licensed Apache-2.0. It adds 40 tokens to every session and 1,187 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 agents, from other repositories
review-e2e
An end-to-end (E2E) test review agent checks whether tests correctly reproduce real user journeys through an application.
mobile-impl
A Japanese-language agent for designing, implementing, and end-to-end testing mobile applications. It plans screen and state flows, adds identifiers for automated testing, and forbids fake implementations or skipped validation and security.
report-builder
Specialized agent for organizing review reports and evidence. Used when executing the /done skill or when report creation is needed.
backend-impl
A backend development agent for building and checking web APIs and their databases. It follows test-first development (writing tests before implementation) and requires real framework-based tests.
review-code-security
A review agent that examines code quality and security, including types, error handling, repeated logic, injection risks, access control, and exposed sensitive data. It can also advise on a proposed design before coding begins.
review-ui-ux
A review agent for user interfaces, the parts of software people see and operate. It checks accessibility, keyboard use, design consistency, wording, and responsive behavior, or gives design advice before implementation.