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 commands/everywan-dev/claude-code-engineering/routegit clone --depth 1 https://github.com/everywan-dev/claude-code-engineeringWhat 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 | $0.00014 | $0.00743 |
| Opus 5 | $0.00007 | $0.00371 |
| Sonnet 5 | $0.00003 | $0.00149 |
| Haiku 4.5 | $0.00001 | $0.00074 |
Grade A, and why
route 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.
What it actually says
Decide how much validation the user's change deserves, before they are under pressure to say "not much".
$ARGUMENTS is what they are about to do, in their own words. If it is empty,
describe the current change from the working tree instead.
Follow these steps precisely:
-
Work out which files the change touches. If the user named files, use those. Otherwise run
git diff --name-only HEADandgit status --porcelainand use the paths from both — staged, unstaged and untracked. If the working tree is clean and$ARGUMENTSis empty, ask what they are about to do and stop. -
Run the router, passing every touched file with its own
--path. Prefer the console script; fall back to the module form, which is what works from an installed plugin where the CLI is not on thePATH:validated-memory route "$ARGUMENTS" --path FILE [--path FILE ...] # or PYTHONPATH="${CLAUDE_PLUGIN_ROOT}" python3 -m validated_memory route "$ARGUMENTS" --path FILE [...] -
Report its answer verbatim first. The level, the matched signals, the model, the effort and the agents. Do not paraphrase and do not soften it. The value of this command is that the reasoning is visible, so the user can disagree with a specific signal rather than with a verdict.
-
If it reports that no signal matched, say so plainly and explain that the level 2 answer is the safe direction rather than a reading of their change. Then ask whether it touches authentication, permissions, money, customer data, migrations, network rules or certificates — and if it does, re-run with that said explicitly in the description.
-
Do not argue the level down. If the user thinks it is too high, the way to change the answer is to change the description or the signal list, both of which are visible and editable. Negotiating the number in conversation is the exact failure this command exists to prevent.
-
Then say what happens next, concretely:
- Level 1 — one review. Name which of the two agents fits.
- Level 2 — two independent reviews. State plainly that the agent or person who implemented it is not one of the two, and that reviewing one's own work a second time does not count.
- Level 3 — three, and one of them has to actively try to break the change. Approving is not the devil's advocate's job.
-
Offer the matching skill for what they are about to do, using
pick-the-right-skillif nothing obvious fits. For level 3 specifically,map-the-attack-surfaceandwrite-the-rollback-plan-firstare almost always worth reading before starting.
Keep the whole response short. This command answers one question; it is not a review, and it never gates anything — it exits 0 by design.
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.
- 2d ago First seen · 62 lines · 14 tokens per session scan A 582dc24a2117
route is a command published in the GitHub repository everywan-dev/claude-code-engineering (2 stars, last pushed 14d ago), licensed Apache-2.0. It adds 14 tokens to every session and 743 once invoked, about $0.0001 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 commands, from other repositories
review-renovate
Review and merge renovate PRs with automerge configuration updates.
plan-regression-tests
Plan regression tests for existing code with it.skip statements.
ado-pull
Pull latest changes from Azure DevOps (like git pull). Supports increment, project, or full living docs sync.
github-push
Push local progress to GitHub Issues (like git push). Updates issue with task completion checklist.
export
Export the wiki (or architecture model) to markdown, HTML, JSON, or Structurizr DSL.
issue-amend
Re-snapshot the active issue's scope from the spec, clear verified and reviewed receipts, and record the change as a permanent amendment.