Bootstrap a new session to continue working on an existing plan. Reads plan artifacts, builds context, validates prerequisites, and prepares the agent for implementation. Use this skill at the start of a session when continuing a multi-session plan.
Bridges the gap between planning artifacts and implementation. When a new session starts and the agent needs to continue work on an existing plan, this skill:
Planning skills (create-plan, update-plan) produce artifacts. Implementation happens across sessions. Without a defined bootstrap procedure, every new session starts with ad-hoc exploration -- the agent must independently discover which plan exists, which phase is active, what was decided, and what files to read. This wastes context budget and risks missing critical decisions from prior sessions.
This skill codifies the read-path that complements the write-path of the other planning skills.
Do NOT use this skill to create or modify plans -- use create-plan or update-plan instead.
doc-explorer: optionally, to validate prerequisites that require codebase inspection (e.g., "Phase 1 tests must pass"). Results are written to docs/, not returned as text.doc-explorer: this skill reads artifacts, it does not write them.Find active plans:
plans/ directory for existing plan directoriesplan.md to check status (active or draft)question tool to ask the user which plan to resumecreate-planRead artifacts in this specific order -- each builds on the previous:
plans/<name>/plan.md -- Understand objective, requirements, DoD, phase overviewplans/<name>/todo.md -- Identify the active phase, current task state, and Phase Context linksplans/<name>/handovers/ (most recent by filename) -- Session context, decisions, blockers, next steps from the last sessionplans/<name>/phases/phase-N.md -- What/Why for the current phase, acceptance criteriaplans/<name>/implementation/phase-N-impl.md -- How to implement, Required Context files, implementation stepsThis order is intentional:
From the implementation plan's Required Context section, read each listed file:
docs/modules/<name>.md) -- understand the code being modifieddocs/features/<name>.md) -- understand the feature being builtIf Required Context is empty or the section doesn't exist, check the Affected Modules table for module doc links and read those instead.
If no project documentation exists under docs/, note this as a gap and suggest running generate-docs.
Check the phase document's Prerequisites section:
doc-explorer via Task toolquestion tool to confirm with the userIf prerequisites are not met, inform the user and stop. Do not proceed with implementation on unmet prerequisites.
Summarize what was gathered into a concise briefing for the session. This is NOT written to a file -- it is presented in the conversation to establish shared context:
Briefing structure:
Use the question tool to confirm with the user:
After confirmation, the session is bootstrapped. The agent can begin implementation, using update-plan to track progress as work proceeds.
| Skill | Relationship |
|---|---|
create-plan | Creates the artifacts that resume-plan reads. Must run before resume-plan is useful. |
update-plan | Used DURING implementation (after resume-plan bootstraps context) to track progress. |
generate-handover | Creates the handover that resume-plan reads at session start. |
generate-docs | Creates module/feature docs referenced by implementation plans' Required Context. |
update-docs | Should run after implementation to keep docs in sync with code changes. |
plans/. It is ephemeral session context.docs/ doesn't exist, suggest generate-docs. If no handover exists, note it (first session on this plan -- that's fine).explore subagent type.question tool: For plan selection, prerequisite confirmation, and briefing sign-off.