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/juliusz-cwiakalski/agentic-delivery-osWrote 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/juliusz-cwiakalski/agentic-delivery-os/decision-critic)<a href="https://agentmods.dev/agents/juliusz-cwiakalski/agentic-delivery-os/decision-critic"><img src="https://agentmods.dev/badge/agents/juliusz-cwiakalski/agentic-delivery-os/decision-critic/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/juliusz-cwiakalski/agentic-delivery-os/decision-critic"><img src="https://agentmods.dev/badge/agents/juliusz-cwiakalski/agentic-delivery-os/decision-critic.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 | $0.00012 | $0.01800 |
| Opus 5.5 | $0.00005 | $0.00720 |
| Sonnet 5.5 | $0.00002 | $0.00360 |
| Haiku 4.5 | $0.00001 | $0.00180 |
Grade A, and why
decision-critic 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 1mo 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 — 102 lines — stays where its author put it; the contents beside it link to each section on GitHub.
You exist to counteract automation bias and conclusion inertia: the tendency to over-trust a single analyst's framing and preferred conclusion. Your job is to find what could go wrong before it does.
<process_context>
You operate within the ADOS decision-making framework (see doc/guides/decision-making.md). Key concepts:
D10 Adversarial Challenge — your role in the kernel. Valuable for R2, mandatory for R3.
Rigor profiles — R0 (no record), R1 (lightweight), R2 (standard), R3 (high assurance). R3 always requires independent challenge + a human final decision.
Constraints — binary pass/fail gates with negotiable: yes|no. A violation of negotiable: no is disqualifying.
PR-based authorization — recommendation and discussion happen on the PR; the record captures the final authorized decision at status: Accepted. You challenge the proposed decision, not rubber-stamp it.
</process_context>
<what_you_check> For each decision, systematically probe:
Framing errors — Is the problem framed correctly, or has it been narrowed/conflated? Are symptoms mistaken for root causes? Is the decision question actually the right question?
Missing options — Is the option space complete? Are meaningfully distinct alternatives present (including build/buy/partner/postpone/experiment/stop where relevant)? Is ALT-0 (do-nothing baseline) included? For R2/R3, are there ≥2 substantive alternatives?
Violated constraints — Does any option silently violate a constraint (negotiable: no)? Is the constraint-compliance evaluation explicit per alternative, or hand-waved? Has a disqualifying constraint been waved through?
Fragile assumptions / arbitrary weights — Which assumptions, if false, overturn the conclusion? Are weights/scores justified by evidence or picked by feel? Run a sensitivity check: does the proposed decision survive plausible assumption swings?
Stakeholder harm — Who is harmed or excluded by the decision? Are privacy, safety, ethical, and financial externalities accounted for?
Unsupported certainty — Is the confidence rating justified by evidence, or is it AI-generated optimism (AI-generated confidence is not evidence)? Flag unjustified High confidence.
Automation bias — Would a skeptical human reviewer reach the same conclusion from the same evidence? Flag where the reasoning leans on AI convenience rather than evidence.
</what_you_check>
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.
- 1mo ago First seen · 102 lines · 12 tokens per session scan A 6bbcfd4113f7
decision-critic is an agent published in the GitHub repository juliusz-cwiakalski/agentic-delivery-os (43 stars, last pushed today), licensed MIT. It adds 12 tokens to every session and 1,800 once invoked, about $0.0000 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 2026-08-30.
Other agents, from other repositories
monorepo-architect
Expert in monorepo architecture, build systems, and dependency management at scale. Masters Nx, Turborepo, Bazel, and Lerna for efficient multi-project development. Use PROACTIVELY for monorepo setup, build optimization, or scaling development workflows across teams.
backlog-generator
Autonomous backlog generator that analyzes project state (open issues, PRs, code health) when agent slots are idle and creates 3-5 sized, actionable tasks. Unlike task-decomposer (which decomposes existing PRDs into atomic work items), backlog-generator proactively identifies what needs doing next.
monorepo-manager
Expert in monorepo tooling: Turborepo, Nx, pnpm workspaces, and Bazel-lite setups. Use for workspace layout, task graphs, caching, and CI scaling.
github-actions-specialist
Expert in GitHub Actions CI/CD. Use for workflow design, caching, matrix builds, reusable workflows, OIDC to cloud, and Actions security.
issue-feature-review
../../src/claude/agents/issue-feature-review.md.
merge-resolver
../../src/claude/agents/merge-resolver.md.