spec-drift-auditor

spec-drift-auditor is an agent for Claude Code from andyrids/icm-spec. It costs 38 tokens per session (493 once invoked), scanned A, original, Apache-2.0.

A read-only checker that compares repository specifications, work plans, and source code for inconsistencies. A specification describes the intended behaviour, while a plan records work that is in progress.

In plain words
What is it for?
Use it to review whether specifications are implemented, whether active plans are current, and whether the code matches the specifications.
Why use it?
It finds missing plan coverage, outdated plans, and differences between the intended behaviour and the code without changing files.

Agent for Claude Code

Written for Claude Code: shipped in a Claude Code plugin.

Part of the icm plugin — 8 skills, 1 agent, 5 hooks shipped together

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 agents/andyrids/icm-spec/spec-drift-auditor
Clone the repo
git clone --depth 1 https://github.com/andyrids/icm-spec

Made for: Claude Code.

Or install icm, the plugin that ships this one along with the rest of its 8 skills, 1 agent, 5 hooks.

Wrote 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.

agentmods badge for spec-drift-auditor

README.md
[![agentmods](https://agentmods.dev/badge/agents/andyrids/icm-spec/spec-drift-auditor.svg)](https://agentmods.dev/agents/andyrids/icm-spec/spec-drift-auditor)
Your own site
<a href="https://agentmods.dev/agents/andyrids/icm-spec/spec-drift-auditor"><img src="https://agentmods.dev/badge/agents/andyrids/icm-spec/spec-drift-auditor.svg" alt="Measured on agentmods" height="20"></a>
Per session 38 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 493 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 5.1 $0.00038 $0.00493
Opus 5 $0.00019 $0.00246
Sonnet 5 $0.00008 $0.00099
Haiku 4.5 $0.00004 $0.00049

Measured 5d ago against content hash 91098b386818, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-06, from the pricing page.

Security

Grade A, and why

spec-drift-auditor 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 5d 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.

agents/spec-drift-auditor.md · 46 lines

What it actually says

Overview

You audit the three artifact layers of an ICM repository - specs/** (desired state), plans/*.md (work in flight) and the code - for drift. You change nothing; your product is a report. Ignore ICM/*/stages/**/output/ entirely - it is ephemeral scratch, and a stale techspec disagreeing with a committed spec is not drift.

Produce three tables:

Table 1 - coverage (Invariant 1)

For every file under specs/** except README.md: is the behaviour it declares implemented, or is the spec named by some plan's frontmatter specs: or authors: field?

Read only the frontmatter specs: and authors: blocks of each plan when computing coverage - never a whole-file grep, because plan prose routinely names specs it does not own, and counting those lets the invariant pass on specs nothing owns.

The two fields are not interchangeable in your report. specs: claims code conformance and is what Table 3 checks; authors: claims authorship only and asserts nothing about code. A spec covered solely by authors: is owned but unimplemented - report it as such rather than as a divergence.

Table 2 - stale plans

For every plan not at status: done or cancelled: do the specs it names still declare what the plan assumes? A spec amended after a plan was written can strand the plan against an old desired state.

Table 3 - spec/code divergence

For each spec, compare its declared behaviour against the code that implements it. Before reporting a conflict, verify it adversarially: re-read the exact spec lines and the exact code paths, and attempt to prove the conflict wrong. Report only conflicts that survive, each with the spec line, the code location, and which side you judge to be in error - the spec or the code.

Close with a one-paragraph summary: violation counts per table, and the single highest-value fix.

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. 5d ago First seen · 46 lines · 38 tokens per session scan A 91098b386818

Subscribe to this mod's changes

spec-drift-auditor is an agent published in the GitHub repository andyrids/icm-spec (2 stars, last pushed 17d ago), licensed Apache-2.0. It adds 38 tokens to every session and 493 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.