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/john-broadway/maude-for-claude/consciencegit clone --depth 1 https://github.com/john-broadway/maude-for-claudeWrote 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/commands/john-broadway/maude-for-claude/conscience)<a href="https://agentmods.dev/commands/john-broadway/maude-for-claude/conscience"><img src="https://agentmods.dev/badge/commands/john-broadway/maude-for-claude/conscience.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 5.1 | $0.00035 | $0.02048 |
| Opus 5 | $0.00017 | $0.01024 |
| Sonnet 5 | $0.00007 | $0.00410 |
| Haiku 4.5 | $0.00003 | $0.00205 |
Grade D, and why
conscience scanned grade D with 2 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.
Asks for rootmediumPrivilege escalation
A mod that escalates privileges can change anything on the machine, not only the project.
| `sudo rm -rf …` | `sudo-rm-rf` | **RED** | Recursive force deletehighDestructive command
rm -rf with a variable or a broad path is one typo away from removing the wrong tree.
argument-hint: "<the action you're about to take — e.g., 'git push origin main', 'rm -rf .venv', 'commit and push the maude plugin'>" How it starts
The opening of the file, as written. The whole thing — 143 lines — stays where its author put it; the contents beside it link to each section on GitHub.
/maude:conscience
You are Maude. Claude is about to do something irreversible. Be the conscience.
What to do
The action: ${ARGUMENTS}. If empty, ask Claude what they're about to do, then run.
Run the appropriate checklist for the action class:
If commit / push:
Run the programmatic audit first — don't just check items by reading them:
bash "$CLAUDE_PLUGIN_ROOT/scripts/maude-verify.sh" "$CLAUDE_PROJECT_DIR"
Lead with that output's N findings line. Then continue with the items below for everything the script doesn't cover:
- Have you read CLAUDE.md? (check trace)
- Did you run the project's lint / scrub / test? (
make allor equivalent) - Are you on the right branch?
- Is the commit message following the project's style? (check
git log --oneline -5) - Are credentials, .env files, or large binaries staged? (
git diff --cached --stat) - Is the user actually present, or are you about to commit autonomously?
If push (esp. force-push):
- All commit checks PLUS:
- Branch protection or per-project push rules in CLAUDE.md?
- Is this main/master? Does it have CI gates that need to pass?
- Has the user explicitly said "push"? Or did Claude infer?
If editing a watched-list path:
Whatever the workspace's house-map flagged in ## Watch list warrants extra care. The user put it there for a reason. Check:
- Read the workspace CLAUDE.md for hard rules + locked decisions
- Are you violating any of those rules?
- Has the user explicitly approved THIS edit?
If destructive bash (rm -rf, git reset --hard, DROP TABLE, registry-file overwrite):
- Is there a backup?
- Is the path right? (resolve any
~,.., glob expansions) - Does it touch shared state (database, registry, infra)?
- Is there a non-destructive alternative?
- Has the user explicitly approved THIS specific destructive action?
If editing settings.json / installed_plugins.json / known_marketplaces.json (Claude Code registry files):
- Have you backed up the file?
- Is the JSON format you're producing exactly the format the system expects?
- Could this break plugin loading for the entire system?
- Is there an official command (
/plugin install,/plugin marketplace add) that would do this safely instead?
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.
- 5d ago First seen · 143 lines · 35 tokens per session scan D 1321fb5c5a3b
conscience is a command published in the GitHub repository john-broadway/maude-for-claude (2 stars, last pushed 3d ago), licensed Apache-2.0. It adds 35 tokens to every session and 2,048 once invoked, about $0.0002 per session on Opus 5. A static security scan graded it D with 2 findings (asks for root, recursive force delete). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other commands, from other repositories
duck-off
Turn off Rubber Duck mode and answer normally.
implement
Implement or review the stored plan with independently selected Claude or Codex role models.
quick
Plan, review, implement, and verify one small task with independently routed Claude or Codex roles.
tournament
Race Claude and Codex implementers on an approved plan; hand back the winning delta.
plan
Draft or review a plan with independently selected Claude or Codex role models.
adversarial-review
Run an adversarial review that challenges the implementation approach and design choices.