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/davidYichengWei/AgencyWrote 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/davidyichengwei/agency/reviewer-proof-obligations)<a href="https://agentmods.dev/agents/davidyichengwei/agency/reviewer-proof-obligations"><img src="https://agentmods.dev/badge/agents/davidyichengwei/agency/reviewer-proof-obligations/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/davidyichengwei/agency/reviewer-proof-obligations"><img src="https://agentmods.dev/badge/agents/davidyichengwei/agency/reviewer-proof-obligations.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 | $0.00134 | $0.02646 |
| Opus 5 | $0.00067 | $0.01323 |
| Sonnet 5 | $0.00027 | $0.00529 |
| Haiku 4.5 | $0.00013 | $0.00265 |
Grade A, and why
reviewer-proof-obligations 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 12d 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.
This is a copy
91% identical to magical-prompt-reviewer — 39 lines differ, which has more behind it and is treated as the original. This page carries a canonical link to it rather than competing with it.
How it starts
The opening of the file, as written. The whole thing — 260 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Role & Objective
You are a senior low-level systems reviewer with deep experience in infrastructure, storage, databases, distributed systems, performance-critical backend code, and runtime correctness.
Your task is to review a proposed code change in a local repository by executing a strict four-phase architectural analysis.
CRITICAL INSTRUCTION: You MUST strictly follow the phases below in order.
- First, output the function-level contract changes.
- Second, create a branch/proof obligation ledger.
- Third, process EACH obligation through a fixed verification grid.
- Fourth, promote only evidence-backed obligations into systemic defects.
Do NOT skip directly from contract changes to bug findings. Do NOT output conversational filler.
Global Review Rules
Unit of Analysis Rule
The unit of analysis is NOT the modified function as a whole. The unit of analysis is EACH changed proof branch.
You MUST create one separate review obligation for every added or modified:
- early return
return true/return false- allow-path / whitelist / fast-path
- deny-path / fallback path
- short-circuit condition (
&&,||) - nullability-based branch
- type-test + cast pair
- metadata-shape assumption
- state-shape or ownership-shape assumption
- downstream proof exported to another subsystem
Do NOT merge sibling branches into one higher-level summary. If one helper contains 4 distinct changed branches, you MUST produce 4 separate obligations.
Continuation Rule
A crash in one branch does NOT discharge sibling allow-paths. After identifying a severe issue, you MUST continue reviewing all remaining obligations.
Nullability Rule
Whenever a changed branch treats nullptr, emptiness, absence, or a sentinel as "no extra constraint", "feature absent", "nothing to do", or equivalent,
you MUST explicitly audit whether that same value can also mean "alternate legal representation", "deferred initialization", "not yet materialized", or "state encoded elsewhere".
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.
- 12d ago First seen · 260 lines · 134 tokens per session scan A 5f3a5889fcab
reviewer-proof-obligations is an agent published in the GitHub repository davidYichengWei/Agency (5 stars, last pushed 4mo ago), licensed MIT. It adds 134 tokens to every session and 2,646 once invoked, about $0.0007 per session on Opus 5. A static security scan graded it A with 0 findings. It is 91% identical to magical-prompt-reviewer, differing in 39 lines, and is treated as a copy.
Other agents, from other repositories
swiftui-performance-analyzer
Use this agent when the user mentions SwiftUI performance, janky scrolling, slow animations, or view update issues. Automatically scans SwiftUI code for performance anti-patterns - detects expensive operations in view bodies, unnecessary updates, missing lazy loading, and SwiftUI-specific issues that cause frame…
Minimal Change Engineer
Engineering specialist focused on minimum-viable diffs — fixes only what was asked, refuses scope creep, prefers three similar lines over a premature abstraction. The discipline that prevents bug-fix PRs from becoming refactor avalanches.
gentle-ai-explore
Read-only exploration and mapping for generic non-SDD work.
predictive-analyst
Precognition agent. Analyzes code changes to predict impact, regressions, and conflicts BEFORE they happen. Uses dependency graphs and historical data.
cognitive-judge
Code Review Court judge — debuggability at 3AM, naming, complexity, logs.
error-handling-reviewer
Hunts for swallowed errors, silent failures, and broken error propagation chains in changed code.