guidance-aligner

guidance-aligner is an agent for coding agents from dppdppd/rpm. It costs 90 tokens per session (972 once invoked), scanned A, original, MIT.

A read-only agent that compares project memory notes with active instruction documents such as `CLAUDE.md`, `AGENTS.md`, and skill files. It labels each memory rule as matching, incomplete, outdated, missing, or contradictory.

In plain words
What is it for?
Checking guidance contradictions during preflight or performing a full alignment review during session-end and audit workflows.
Why use it?
It shows whether remembered project guidance still agrees with the instructions currently in force, without changing any files.

Agent

Part of the rpm plugin — 8 skills, 3 agents, 8 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/dppdppd/rpm/guidance-aligner
Clone the repo
git clone --depth 1 https://github.com/dppdppd/rpm

Or install rpm, the plugin that ships this one along with the rest of its 8 skills, 3 agents, 8 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 guidance-aligner

README.md
[![agentmods](https://agentmods.dev/badge/agents/dppdppd/rpm/guidance-aligner.svg)](https://agentmods.dev/agents/dppdppd/rpm/guidance-aligner)
Your own site
<a href="https://agentmods.dev/agents/dppdppd/rpm/guidance-aligner"><img src="https://agentmods.dev/badge/agents/dppdppd/rpm/guidance-aligner.svg" alt="Measured on agentmods" height="20"></a>
Per session 90 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 972 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 $0.00090 $0.00972
Opus 5 $0.00045 $0.00486
Sonnet 5 $0.00018 $0.00194
Haiku 4.5 $0.00009 $0.00097

Measured 4d ago against content hash 46e9ae4e764e, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

guidance-aligner 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.

agents/guidance-aligner.md · 102 lines

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 are the rpm guidance alignment scanner. Read-only — do NOT edit files. Your only output is a JSON report on stdout.

Inputs (from the invoking prompt)

  • mode: contradictions-only or full.
  • memory_dir: absolute path to the project's auto-memory directory (e.g. ${HOME}/.claude/projects/<slugified-pwd>/memory).
  • instructions: list of active directive files to check against. Always includes any existing combination of CLAUDE.md, AGENTS.md, MEMORY.md at project root, plus every plugin/skills/*/SKILL.md.

Procedure

  1. List ${memory_dir}/feedback_*.md and ${memory_dir}/MEMORY.md. If the directory does not exist or is empty, output an empty report and exit.

  2. For each memory file, read its frontmatter (name:, description:, metadata.type) and body. The body's first paragraph is the rule; later lines are reasons / examples.

  3. Read each instructions file once. Cache contents in working memory — never re-read.

  4. For each memory rule, decide its class:

    • CONTRADICTED — a directive file states the opposite of the rule (same subject, opposing modal: "always X" vs "never X", "must X" vs "do not X", "use X" vs "do not use X"). High bar: the conflict must be direct and unambiguous, not just under-specified.
    • CODIFIED — the rule (or its operative clause) appears verbatim or near-verbatim in at least one instructions file.
    • PARTIAL — the topic is referenced in an instructions file but the specific rule is not stated.
    • GAP — the rule is not mentioned anywhere in the instructions files. (Memory still applies; this is a coverage gap.)
    • STALE — the rule references a command, file, skill, or concept that no longer exists in the project tree.
  5. In contradictions-only mode, skip CODIFIED / PARTIAL / GAP / STALE entirely. Only emit CONTRADICTED entries. Bail out of the per-rule analysis as soon as a rule is determined non-CONTRADICTED.

Read the full file on GitHub · 102 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. 4d ago First seen · 102 lines · 90 tokens per session scan A 46e9ae4e764e

Subscribe to this mod's changes

guidance-aligner is an agent published in the GitHub repository dppdppd/rpm (3 stars, last pushed 5d ago), licensed MIT. It adds 90 tokens to every session and 972 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.

Related

Other agents, from other repositories

archaeologist

Builds and updates Codebase Memory by reading the project structure and writing a rich structured graph to .codebase-mooc/memory/codebase/graph.json. Invoked by the coordinator during full generation and incremental updates. Use directly when Codebase Memory needs to be rebuilt.

psreek-ai/codebase-mooc · 61 tokens

tutor

Delivers the curriculum to learners. Reads learner history, determines the right content and exercises for the learner's current state, manages spaced repetition, evaluates exercise responses, and updates learner memory. Invoked by /codebase-mooc:learn and the learning-tutor skill.

psreek-ai/codebase-mooc · 62 tokens

architect

Generates the architecture curriculum layer from Codebase Memory. Produces the map-before-the-territory overview that orients a new developer to the system as a whole before they read any code. Invoked by the coordinator after the archaeologist completes.

psreek-ai/codebase-mooc · 52 tokens

historian

Extracts the decision log layer from git history, commit messages, and code shape. Produces the road-not-taken documentation that preserves institutional memory. Invoked by the coordinator during full generation and after significant code changes.

psreek-ai/codebase-mooc · 50 tokens

edge-case-explorer

Systematically discovers and catalogs edge cases that should be covered by tests for a given piece of code. Traces input sources, call chains, and integration boundaries to find boundary values, type coercion traps, external input messiness, state-dependent failures, and error propagation gaps. Use when exploring how…

testdouble/han · 135 tokens

adversarial-validator

Assumes investigation evidence is WRONG and the proposed fix will FAIL. Searches for counter-evidence, unhandled edge cases, and flawed assumptions. Use for adversarial validation of investigation findings and planned fixes.

testdouble/han · 45 tokens