Moses Roles
PassAudited by ClawScan on May 1, 2026.
Overview
This instruction-only governance skill is coherent with its stated role-sequencing purpose, but it can persistently shape agent behavior and rely on a local governance file and companion audit script.
Install this skill if you want persistent multi-agent role governance and enforced response sequencing. Before using it, review any AGENTS.md changes, protect the governance state file, and review the separate moses-governance audit script if you install that companion bundle.
Findings (6)
Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.
The agent may prioritize the role hierarchy and sequencing rules over normal free-form responses.
The skill intentionally imposes mandatory response ordering and blocking behavior across agents.
This order is not a suggestion. It is enforced.
Install only if you want this role-governance behavior, and use the documented broadcast override or disable the skill if it conflicts with a task.
If the surrounding agent environment grants powerful tools, the Primary role may be treated as the role allowed to use them.
The Primary role is instructed to have broad tool access, though this is disclosed and aligned with the skill's leadership-role design.
Primary — You lead. You respond first. You set the analytical direction. No action is deferred to Secondary or Observer. Full tool access.
Pair this skill with normal tool-approval safeguards for destructive, account-changing, or external actions.
Audit logging depends on another installed skill bundle, so its behavior depends on code outside this instruction-only artifact.
The skill references an executable helper script from a companion moses-governance bundle that is not included in this artifact set.
python3 ~/.openclaw/workspace/skills/moses-governance/scripts/audit_stub.py log --action "sequence_violation"
Review and trust the companion moses-governance bundle before relying on or enabling its audit logging behavior.
Changes to the governance state file can influence future agent behavior and role constraints.
A persistent local governance file is treated as authoritative context for every response.
Before every response, each agent loads: `~/.openclaw/governance/state.json`
Keep the governance state file under user control and review changes to it, especially in shared workspaces.
The role rules may continue affecting future sessions in that workspace after they are added to AGENTS.md.
The skill recommends persistent workspace-level agent instructions for the role hierarchy.
Add to `~/.openclaw/workspace/AGENTS.md`:
Add these overrides only in workspaces where persistent MO§ES role governance is desired, and remove them when no longer needed.
If the optional secret is set, it should be treated like a credential even though the artifact describes it as local-only.
The skill declares an optional sensitive operator secret, although no code or transmission path is shown in the provided artifacts.
MOSES_OPERATOR_SECRET ... required: false ... sensitive: true ... purpose: "Optional local HMAC signing gate. Never transmitted."
Do not set MOSES_OPERATOR_SECRET unless needed, and store it securely if used.