planning-doc
Planning Doc
Create and maintain PLAN.md files tied to git branches. The plan serves two audiences:
- Agent: determine where things are at and continue from where they left off.
- Human: quickly see the plan, changes, and current status.
Plan Steering
- Always read
PLAN.mdbefore making changes. - On resumed sessions, read the latest Status Updates entry and run
git diff --statto detect drift. - If the user says "continue", use the newest status update to determine what changed last and what to do next.
- Treat
PLAN.mdas a resume log, not a task tracker — keep entries short and factual. - Use checkboxes in the Plan section to track progress.
Workflow
- Determine the current branch:
git rev-parse --abbrev-ref HEAD. - If on the default branch (usually
main):- Stop and prompt the user to create and check out a feature branch before proceeding.
- Derive the plan path:
- If the branch matches
^(feat|fix|chore)/(.+)$, use prefix = match 1 and feature = match 2. - Otherwise use prefix =
featand feature = full branch name. - Plan path:
docs/plans/<prefix>/<feature>/PLAN.md.
- If the branch matches
- If PLAN.md does not exist:
- Create directories and create the plan from
references/plan-template.md. - Fill in the Goal and Plan sections based on the user request.
- Create directories and create the plan from
- If PLAN.md exists:
- Do not overwrite. Update checkboxes, append status updates.
Updating PLAN.md
- After meaningful work, append a Status Updates entry (newest first):
- Change: what changed
- State now: what's true after this change
- Validate: one command to verify
- Check off completed items in the Plan section.
- Add Decisions entries for notable tradeoffs.
- Add Discoveries entries for non-obvious findings or gotchas.
Error Protocol
- Three-strike escalation for recurring failures:
- Diagnose and apply a targeted fix.
- Try a different approach (do not repeat the same failing action).
- Re-check assumptions and update the plan.
- After three failed attempts, escalate to the user with what was tried and the best next options.
Validation
- Before claiming completion, run the
Validate:command from the most recent status update.
Resource
- Planning template:
references/plan-template.md
More from chandima/opencode-config
github-ops
|
2context7-docs
|
2mcporter
Direct MCP access via MCPorter. Use to discover MCP servers, list MCP tools, or call an MCP tool (e.g., chrome-devtools screenshot, firecrawl scrape) when no specific skill exists. Supports any configured MCP server.
1asu-discover
Semantic search across 760+ ASU GitHub repositories via RAG. Use for finding code patterns, integrations, SDKs, and ASU-specific conventions. Domains: PeopleSoft, EDNA, DPL, Terraform, EEL, Vault, CI/CD. Use BEFORE starting ASU integration tasks.
1skill-evals-run
Run the OpenCode skill-loading eval suite for this repo. Use when asked to run skill evals, skill-loading evals, or the skill-evals-run command.
1agent-browser
Browser automation CLI for AI agents. Use when the user needs to interact with websites, including navigating pages, filling forms, clicking buttons, taking screenshots, extracting data, testing web apps, or automating any browser task. Triggers include requests to "open a website", "fill out a form", "click a button", "take a screenshot", "scrape data from a page", "test this web app", "login to a site", "automate browser actions", or any task requiring programmatic web interaction.
1