patterns/mktsk/skills/mktsk/SKILL.md
Use when: converting TODO plan or open GitHub issues into deterministic execution checklist
npx skillsauth add ryderfreeman4logos/cli-sub-agent mktskInstall this skill globally with one command. Works with Claude Code, Cursor, and Windsurf.
3 of 9 scanners reported clean
Some scanners were skipped, did not run, or reported a non-clean status. Review each row below.
mktsk MUST be executed by the main agent/orchestrator.
Do NOT delegate mktsk bookkeeping to csa plan run --pattern mktsk or
csa run --skill mktsk.
Why: mktsk owns progress tracking across auto-compaction. CSA subprocesses cannot update the caller's task list, so delegating mktsk bookkeeping makes task persistence impossible. Implementation work for individual registered items may still be delegated when the pattern's executor policy says so.
Use the host agent's native task-list tool. The pattern uses Claude Code tool names as operation names; adapt them instead of assuming Claude-only tools:
| Host | Create task | Update task | List/resume |
|------|-------------|-------------|-------------|
| Claude Code | TaskCreate | TaskUpdate | TaskGet / TaskList |
| Hermes | todo tool | todo tool | todo tool |
| Codex | update_plan | update_plan | current plan state |
Hermes agents exposed through ~/.hermes/skills must use the todo tool as
the TaskCreate / TaskUpdate / TaskList equivalent. Codex and Hermes must
translate task operations to their host tools rather than inventing or invoking
Claude-only tool names.
If the current host has no durable task-list equivalent, stop and ask the user which tracking tool to use before executing mktsk.
Keep the source of truth in this repository:
patterns/mktsk/skills/mktsk
Expose it to Hermes with a symlink from the Hermes skill directory:
~/.hermes/skills/software-development/mktsk -> patterns/mktsk/skills/mktsk
Do not copy or maintain a separate ~/.hermes version of this skill.
Read the pattern at ../../PATTERN.md (relative to this SKILL.md) and follow it
step by step. You are executing directly — every step runs in your context.
Modes:
path=..., timestamp=..., or the latest mktd output.--from-issues: read open GitHub issues and register one ordered task-list
entry per issue.For csa commands within pattern steps (e.g., csa review --diff), add
--sa-mode true when operating under SA mode.
This is the while-waiting checklist. When you background a csa session wait via run_in_background: true, the next task-notification wakes you up automatically. Do not add manual sleep, polling, a redundant ScheduleWakeup, or /loop on top.
Safe parallel work:
gh pr create yet.Do NOT:
csa run or csa review sessions that could race on git branch or checkout state with the waiting one (single-checkout sequential rule, AGENTS.md 028).git commit, git checkout <other-branch>, or git push.If there is no useful parallel work available, return control and wait for the notification. Do not invent speculative work just to stay busy.
| Command | Effect |
|---------|--------|
| /mktsk | Execute the most recent TODO plan for the current branch |
| /mktsk path=./plans/feature.md | Execute tasks from a specific plan file |
| /mktsk timestamp=01JK... | Execute tasks from a csa todo by timestamp |
| /mktsk --from-issues | Create an ordered task-list backlog from all open issues |
mktd (provides TODO plan), commit (per-task commit workflow)csa-review (per-task review), security-audit (via commit skill)CSA_SKIP_PUBLISH=true), publish steps are skipped.DONE WHEN conditions.CSA_SKIP_PUBLISH=true.)tools
--- name = "csa-setup" version = "0.1.0" description = "Install CSA, Weave, and coding workflow patterns for AI agent bots" homepage = "https://github.com/RyderFreeman4Logos/cli-sub-agent" target = "agent-bots" --- # CSA Setup > Installation guide for AI agent bots (openclaw, Moltis, etc.) to set up > CSA and Weave with optional coding workflow patterns. ## Skill Files | File | Description | |------|-------------| | **skill.md** (this file) | Installation and setup guide | | **skills/AGENTS.
development
Use when: full dev cycle branch->plan->implement->review->PR->merge
development
Use when: running CSA-driven code review, independent model selection
development
Use when running a non-blocking CSA background code health scan that uses csa health and csa tokuin estimate to propose refactoring GitHub issues for files over token or complexity thresholds.