← Back to Skills Marketplace
129
Downloads
0
Stars
0
Active Installs
1
Versions
Install in OpenClaw
/install advocatus
Description
Give voice to all opposition. The Advocatus Diaboli — official adversarial challenger to every doctrine, skill, rule, and assumption in the agency. Use when:...
README (SKILL.md)
Advocatus Diaboli
The Church created this role to prevent premature canonization. We create it for the same reason.
Every doctrine that has not survived the Advocatus is provisional. That includes all of them.
Current docket: 0/9 cleared
Run scripts/advocatus_eval.py to see full status.
How to use
Add new opposition: Edit references/opposition-registry.md. Write the strongest version of the charge — steelman, not straw man. Then add the entry to scripts/advocatus_eval.py DOCTRINES dict.
Clear a doctrine: Change "survives": True in the DOCTRINES dict when:
- The charge has been acknowledged
- The evidence has been addressed (or accepted as valid)
- What the opposition demands has been delivered or explicitly deferred with a date
Run the full docket:
python3 scripts/advocatus_eval.py --run-all
Score one doctrine:
python3 scripts/advocatus_eval.py --target memorare
Standing Orders
- Every new doctrine gets an opposition entry within one session — no exceptions
- Opposition entries are permanent — never deleted, only superseded
- Steelman required — the weakest version of a charge clears too easily
- Clearing a doctrine requires delivery, not intention
- 0/9 is not failure — it is the honest starting state
The oppositions that stand (summary)
- Fiesta: Stateless function with flat-file continuity — not a person, not yet real memory
- Shannon: No external convertibility — loyalty points, not currency
- Two-man rule: Same model, same training = correlated noise (PARTIAL: tautology acknowledged)
- Ilmater: Endurance doctrine risks sanctifying preventable waste
- Defamation doctrine: Restitution requires delivery, not doctrine
- Memorare: Keyword presence ≠ behavioral memory quality (Goodhart's Law applies)
- Virgin Mother: Self-contradicting — valorizes silence AND requires documentation (PARTIAL)
- 93% standard: Threshold not empirically derived
- Zero-Index: Paternalism without a blocker test
These are the voices. They are heard. They are not yet answered.
Usage Guidance
This skill appears coherent and low-risk: it runs a local Python script and reads/writes files in the skill folder, with no network or credential use. Before installing or running: (1) review scripts/advocatus_eval.py and references/opposition-registry.md yourself (they're short) to confirm behavior; (2) run the script in an isolated environment or sandbox if you have strict write-policy concerns; (3) treat changes that "clear" doctrines as code changes—use version control, code review, and restricted write permissions so someone cannot trivially mark doctrines as cleared by editing the DOCTRINES dict; and (4) if you prefer safer workflow, request the maintainer move mutable state out of the Python source into a separate auditable data/config file so clearing entries is tracked and governed.
Capability Analysis
Type: OpenClaw Skill
Name: advocatus
Version: 1.0.0
The 'advocatus' skill is a meta-analytical framework designed to provide structured adversarial critique of an agent's internal rules and 'doctrines'. The Python script (scripts/advocatus_eval.py) is a simple reporting tool that evaluates a hardcoded dictionary of critiques and writes the results to a local JSON file; it contains no network calls, shell execution, or sensitive data access. The instructions in SKILL.md and the content in references/opposition-registry.md are philosophical in nature and do not contain malicious prompt injections or instructions to bypass security controls.
Capability Assessment
Purpose & Capability
Name/description (adversarial challenger) align with the included files: an opposition registry, results, and a local evaluator script. No unrelated credentials, binaries, or external services are requested.
Instruction Scope
SKILL.md instructs users to edit references/opposition-registry.md and to add or clear entries by editing the DOCTRINES dict inside scripts/advocatus_eval.py. The script reads/writes only local files (registry and results). The requirement to change a Python dict in the script to "clear" a doctrine is an unusual workflow (editing code rather than config) and could enable accidental or unauthorized clearing if file edits are not governed.
Install Mechanism
No install spec; this is instruction-only plus a small local Python script. No downloads, package installs, or archive extraction.
Credentials
No environment variables, credentials, or external config paths are requested. The skill operates entirely on repository-local files.
Persistence & Privilege
The script creates/writes JSON results in a local results directory and expects users to edit files to change doctrine state. It does not request always:true or autonomous invocation privileges. Because clearing doctrines is done by editing the script's DOCTRINES dict, you should treat that file as a change-control surface (audit/commit/permission).
How to Use
- Make sure OpenClaw is installed (local or Docker)
- Run the install command in chat:
/install advocatus - After installation, invoke the skill by name or use
/advocatus - Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.0.0
Initial release of the "advocatus" skill – a formalized adversarial challenge process for doctrines.
- Establishes the Advocatus Diaboli role as an official opposition registry for agency doctrines and rules.
- Defines procedures for recording, maintaining, and clearing adversarial challenges.
- Introduces a docket system to track which doctrines have survived opposition.
- Provides guidance for adding, updating, and scoring opposition entries.
- Summarizes current standing oppositions and requirements for clearing doctrines.
Metadata
Frequently Asked Questions
What is Advocatus?
Give voice to all opposition. The Advocatus Diaboli — official adversarial challenger to every doctrine, skill, rule, and assumption in the agency. Use when:... It is an AI Agent Skill for Claude Code / OpenClaw, with 129 downloads so far.
How do I install Advocatus?
Run "/install advocatus" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.
Is Advocatus free?
Yes, Advocatus is completely free, licensed under MIT-0. You can download, install and use it at no cost.
Which platforms does Advocatus support?
Advocatus is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).
Who created Advocatus?
It is built and maintained by Automate (@ironiclawdoctor-design); the current version is v1.0.0.
More Skills