By name: read-task description: Read and understand the current task from the .task/ folder before starting any implementation work. Use this skill at the start of every session to load active task context, understand what's in progress, and identify what needs to be done next.
read-task — Load Task Context from .task/
Folder layout
.task/
*.md # active tasks (top-level = in progress)
completed/ # done tasks — read for historical context only
<feature-name>/ # sub-plans for large features (read all files inside)
Active tasks are .md files at the top level of .task/ (not in completed/).
How to use this skill
At the start of a session (or when resuming work), run:
ls .task/
Then read every active task file using the Read tool (not cat):
Read: .task/<filename>.md
For feature sub-folders, list and then read each file:
ls .task/<feature-name>/
Read: .task/<feature-name>/<file>.md
What to extract from each task file
- Goal — what feature or fix is being built
- Current status — what's done, what's in progress, what's blocked
- Next steps — the concrete next action to take
- Files to touch — components, routes, models affected
Conventions
- Task files use
## Statusor## Progresssections — scan for these first to get a quick read on where things stand - Check
git log --oneline -10to see recent commits related to the task - Cross-reference open PRs with
mcp__github__list_pull_requeststo see if a PR already exists for the task - If a task file has a checklist, items marked
[x]are done;[ ]are pending
Do NOT
- Start writing code before reading the task file
- Assume the task is the same as the last session — always re-read
- Skip sub-folders in
.task/<feature-name>/— they contain the full plan