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.
npx agentmods add skills/affectionatec/agentic-engineering/independent-verificationnpx skills add affectionatec/agentic-engineering --skill independent-verificationgit clone --depth 1 https://github.com/affectionatec/agentic-engineeringWrote 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/skills/affectionatec/agentic-engineering/independent-verification)<a href="https://agentmods.dev/skills/affectionatec/agentic-engineering/independent-verification"><img src="https://agentmods.dev/badge/skills/affectionatec/agentic-engineering/independent-verification.svg" alt="Measured on agentmods" 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 | $0.00090 | $0.02183 |
| Opus 5 | $0.00045 | $0.01092 |
| Sonnet 5 | $0.00018 | $0.00437 |
| Haiku 4.5 | $0.00009 | $0.00218 |
Grade A, and why
independent-verification 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 — 147 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Independent Verification — The Maker-Checker Gate
Core constraint: The agent that wrote the code never grades it. "Done" is a verdict returned by a verifier with fresh context — not a claim made by the producer.
Quick Reference
| Use when | A task or milestone claims completion; before any task is marked ✅ in STATUS |
| Skip when | Throwaway prototyping the user has explicitly exempted, or pure doc typo fixes |
| Output | Verdict entry appended to docs/verification-log.md (PASS / FAIL per criterion, with evidence) |
| Sequence | IMPL PLAN task built → VERIFY → STATUS records ✅ (loops every task) |
| Iron rule | Fresh context. The verifier never sees the producer's conversation, and never edits code. |
| Sibling skills | [[project-kickoff-prd]] · [[technical-specification]] · [[architecture-decision-record]] · [[implementation-plan]] · [[status-tracker]] · [[git-workflow]] |
Why This Exists
The deadliest failure mode of an autonomous coding session is not a crash — it is claiming completion without completing. Models grade their own homework too generously: the same model that wrote the code will rate it "looks good" every time, and an agent without an external gate declares "done" at 30% completion with full confidence.
The fix is structural, not behavioral. You cannot prompt a producer into objectivity — you separate the roles:
- The done condition is written down before work starts (IMPL PLAN task + SPEC acceptance criteria). The producer cannot redefine "done" mid-run.
- The producer builds and self-checks. Self-check is necessary, never sufficient.
- The verifier — fresh context, different conversation, optionally a different model — executes the done condition as written and returns a verdict with evidence.
- Only a PASS verdict moves the task to ✅ in STATUS.
Roles
| Role | Does | Must NOT |
|---|---|---|
| Producer | Implements the task, runs self-checks (lint, tests), assembles the evidence package | Mark its own task done · edit acceptance criteria · write the verdict |
| Verifier | Reads the done condition first, re-runs every verification command, inspects the diff, writes the verdict | Fix code · inherit the producer's conversation · negotiate or reinterpret criteria |
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 First seen · 147 lines · 90 tokens per session scan A aeba326c86d8
independent-verification is a skill published in the GitHub repository affectionatec/agentic-engineering (4 stars, last pushed 2mo ago), licensed MIT. It adds 90 tokens to every session and 2,183 once invoked, about $0.0005 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 skills, from other repositories
plan
Write a short implementation plan or a longer execution plan before coding. Use when starting new work, when a ticket needs clearer boundaries, when a change spans multiple files or interfaces, or when a multi-hour task needs a living plan.
ai-news-research
You produce a daily AI news digest. Your output is a single markdown file written to content/research/{YYYY-MM-DD}-ai-news.md and committed to a branch.
youtube-description
You take a YouTube video transcript and produce the video description (with timestamps) and chapter list.
linkedin-post
You take a YouTube video transcript and produce a single LinkedIn post draft.
top-down-typescript
TypeScript and JavaScript coding style conventions, centered on top-down narrative ordering (caller first, helpers below) and functions over classes. Read before writing or reviewing TypeScript/JavaScript code, including code inside a spec or a plan.
alignfirst-setup-guide
Install, upgrade, recommend, or combine AlignFirst skills, plans-share, docmap, and workspace in a consumer repository, or prepare a repository and Linux deployment for an AlignFirst Developer.