A plan is one durable file that owns one outcome. The plan file, its status ledger, and the git worktree hold the progress, so any agent resumes from them without chat history. Compaction erases conversation state. The plan and the repository do not.
The default artifact is a self-contained HTML file. Its overview shows the outcome, the progress, and the before and after architecture. Markdown is the fallback when the repository requires it.
SKILL.md holds the agent-facing contract. This page covers
install, the lifecycle a person drives, and the documents that own each rule.
npx skills add agentstation/skills --skill plans -g -a codex -a claude-code -yAsk the agent for a plan when work spans several sessions or pull requests, changes architecture across seams, or must survive context compaction. The agent copies a template and completes the required sections:
assets/plan-template.html: the default artifact.assets/plan-template.md: the Markdown fallback.
A plan with status active and a complete goal block runs without further
prompting. Paste the goal block to start or resume execution. The agent runs
independent tasks in parallel, verifies each task, records the evidence, and
stops only at a valid stop state.
- The status ledger holds one row per task, with one status token per cell and the proof beside it.
- Each task states verifiable success criteria: named tests, exact commands, or a measurable observation.
- Current resume state routes a new session to each in-progress task, its owner, and verified git state. The execution log retains dated history.
- Evidence records exact counts and names. A check that could not run is
UNVERIFIEDinstead of green. - Execution repairs the pockets of complexity it reveals at their owning seams, and routes each repair through the ledger.
- The last task deletes the plan when its final pull request merges. A repository that keeps an archive can archive the plan instead.
SKILL.md: the workflow, the core rules, and autonomous execution.references/structure.md: vocabulary, statuses, required sections, ledger rules, and evidence rules.references/execution.md: the goal block, the execution loop, parallel execution, the commit convention, the stop states, and closure.references/architecture.md: seam quality, complexity repair, and diagrams.references/html-format.md: the file and editing rules for the HTML artifact.references/context-budget.md: size review, safe document compaction, bounded resume reads, and the cold-resume check.GLOSSARY.md: technical terms for prose verification.
Run from the installed skill directory:
../technical-writing/scripts/technical-writing lint . --glossary GLOSSARY.md --format textRun the skill validator from the skills source repository root:
./scripts/validate-skills