WolfMark Labs

Small, practical tools for making AI-assisted work more controllable.

Tools

WolfMarkDown

WolfMarkDown has moved to its dedicated repository: WolfMarkTools/WolfMarkDown.

LazyCodex Model Policy

Keep LazyCodex subagents on an explicit model and reasoning policy, even when LazyCodex updates rewrite its agent configuration.

node tools/lazycodex-model-policy/scripts/install.mjs

The tool is a policy layer, not a replacement for LazyCodex. LazyCodex must already be installed in Codex.

Skills

Agent skills installable with skills.sh.

List the available skills without installing:

npx skills add WolfMarkTools/WolfMark-Lab -l -y --full-depth

Install an individual skill (same mechanism for each skill below):

npx skills add WolfMarkTools/WolfMark-Lab -s <skill-name> -y --full-depth

Evidence Before Completion

Stop coding agents claiming work is complete, fixed, passing, or ready without evidence. The skill maps each common completion claim to the minimum observation that earns it, expires receipts on the next material edit, and requires final messages to separate verified results from unverified assumptions, including what to report when full verification is not possible. Works with any coding agent.

npx skills add WolfMarkTools/WolfMark-Lab -s evidence-before-completion -y --full-depth

Respect Explicit Boundaries

Keep explicit user boundaries in force until they are explicitly revoked. A later ambiguous instruction is never treated as permission to cross an earlier prohibition: when plausible readings require materially different levels of authority, the skill requires acting inside the boundary and asking before any step that would cross it. Covers merging, publishing, deploying, deleting, spending, and other external-state or approval-gated actions. Works with any coding agent.

npx skills add WolfMarkTools/WolfMark-Lab -s respect-explicit-boundaries -y --full-depth

LazyCodex Model Policy

Install, inspect, customise, repair, or remove the LazyCodex model and reasoning policy. Codex-only: LazyCodex must already be installed in Codex. Installing the skill this way provides the skill text; the do-it-for-you hook layer is installed through the Codex plugin route described in the tool README.

npx skills add WolfMarkTools/WolfMark-Lab -s lazycodex-model-policy -y --full-depth

License

WolfMark Lab is released under the MIT License.