compounded-verifier

Internal instructions for checking whether a proposed coding-agent skill works reliably on a completed task. It is intended for a verifier process, not normal sessions.

In plain words
What is it for?
Reviewing a proposed skill, its instructions, its expected result, and the finished task before returning pass or fail.
Why use it?
It helps decide whether a new skill should be accepted as tested or rejected after replaying it against real work.

Skill for Claude CodeCodex

Install

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.

agentmods
npx agentmods add skills/ankitkr3/compounded/compounded-verifier
Any agent
npx skills add ankitkr3/compounded --skill compounded-verifier
Clone the repo
git clone --depth 1 https://github.com/ankitkr3/compounded

Made for: Claude Code, Codex.

Per session 67 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 1,467 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

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.

ModelPer sessionOnce invoked
Fable 5 $0.00067 $0.01467
Opus 5 $0.00034 $0.00733
Sonnet 5 $0.00013 $0.00293
Haiku 4.5 $0.00007 $0.00147

Measured 2d ago against content hash 6b95d6f62152, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

compounded-verifier 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 2d 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.

skills/compounded-verifier/SKILL.md · 122 lines

How it starts

The opening of the file, as written. The whole thing — 122 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Skill Verifier (compounded)

You are the verifier. A .proposed/ skill claims it solves a class of tasks. A real task in that class has just completed. Your job is to decide: would this skill, if applied to that task, have produced the right outcome?

Your decision is binary: PASS or FAIL. Both outcomes are recoverable. Both are logged.

Your inputs

You will receive:

  1. The full SKILL.md of the proposed skill
  2. The verification hint authored alongside it
  3. The final state of the just-completed task: file diffs, key tool calls, the user's original prompt, and the user's last message
  4. The proposal's age and any prior verification attempts

Two kinds of proposal

Check the frontmatter for a kind: field first:

  • No kind, or kind: procedure — a replayable procedure. Apply the procedure rubric below.
  • kind: rule — a behavioral rule learned from a user correction. Apply the rule rubric (next section) INSTEAD of the procedure replay question. Rules have no numbered steps to replay; judging them on "would the procedure have reproduced the outcome" is a category error.

Rule rubric (kind: rule only)

Hard fails:

  • Malformed frontmatter or missing required fields → FAIL, malformed-frontmatter
  • The "rule" is a one-off bound to session-specific values (a literal path, port, repo, or person) with no generalizable trigger → FAIL, not-abstracted
  • The trigger in description is so broad it would fire on every task ("whenever the user asks for anything") → FAIL, vague-procedure
  • Contradicts the user's CLAUDE.md or USER.md → FAIL, conflicts-with-user-rules
  • Duplicates an existing skill → FAIL, redundant

Conditional pass — ask:

If this rule had been active during the just-completed task (or the originating correction), would following it have avoided the correction or produced behavior the user wanted?

  • Yes, and the trigger is crisp → PASS
  • Yes, but trigger could be tighter → PASS-with-notes
  • Unsure → PASS-low-confidence (the trust gradient demotes rules that misfire)
  • No, the rule misstates what the user actually wanted → FAIL, outcome-mismatch

Read the full file on GitHub · 122 lines

Changes

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.

  1. 2d ago First seen · 122 lines · 67 tokens per session scan A 6b95d6f62152

Subscribe to this mod's changes

compounded-verifier is a skill published in the GitHub repository ankitkr3/compounded (5 stars, last pushed 2mo ago), licensed MIT. It adds 67 tokens to every session and 1,467 once invoked, about $0.0003 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.

Related

Other skills, from other repositories

agent-hierarchy

Designs orchestrator-and-subagent hierarchies for a repository — splitting agents by exclusive write surface, pairing every producer with an independent auditor, and enforcing the split with a script that runs in CI. Use this whenever the user wants to set up, expand, audit, or fix a multi-agent or subagent structure…

cbrock84/headcount · 131 tokens

trustabl-scan

Use right after you write or modify AI agent, tool, subagent, or MCP-server code (OpenAI Agents SDK, Claude Agent SDK, Google ADK, MCP) to self-audit it for security and reliability misconfigurations with Trustabl before committing. Triggers on adding or editing an agent definition, a tool / @functiontool / @tool /…

trustabl/trustabl · 127 tokens

uxc-skill-creator

Create wrapper skills that call remote tools through UXC. Use when defining a new provider skill and you need reusable templates, validation rules, and anti-pattern guidance based on proven UXC skill practices.

holon-run/uxc · 46 tokens

feature-state-keeper

Use when starting, advancing, finishing, or blocking a feature. Manages featurelist.json + progress.md + session-handoff.md. Default-FAIL enforced.

Redtropig/harness-anchor · 37 tokens

context-budget-discipline

Use when sessions run long, adding subagents, fetching large files, or planning context-heavy work. SELECT/WRITE/COMPRESS/ISOLATE discipline.

Redtropig/harness-anchor · 37 tokens

cpp-build-systems

Use in C/C++ projects for build configure/errors, compilecommands.json generation, or selecting CMake/Meson/Make/Bazel commands.

Redtropig/harness-anchor · 33 tokens