Optional task-management tools for any kind of project. They require Python 3.9+ and Bash or Zsh, but no third-party Python packages. They do not prescribe the project's language, platform, or subject.

Tasks live in docs/tasks.md. Project identity and optional commands live in project-config.json.

Environment

source ./scripts/env.sh

This adds scripts to PATH and defines task aliases. It sources trusted local shell settings from scripts/env.sh.local, or .env if the first file is absent. Both are ignored by Git. An existing TASKS_FILE is preserved. TASK_ID_PREFIX overrides task frontmatter; the default prefix is EX. All task commands accept -f /path/to/tasks.md.

Query and Edit Tasks

./scripts/get-tasks --all
./scripts/get-tasks --accept 'Status:Testing'
./scripts/get-tasks --planning
./scripts/get-tasks --sort asc
./scripts/get-task EX-001
./scripts/get-task 1 --json
./scripts/add-task "Describe the first deliverable" -p example-project -t task -s triage
./scripts/update-task 1 --status ready
./scripts/work-on-task 1 --start
./scripts/update-task 1 --status testing --check-all
./scripts/update-task 1 --append-description "- [ ] Additional review criterion"
./scripts/update-tasks

get-tasks: triage tasks by default; --all includes completed work. Supports column selection and filters; --help lists the options.
get-task: metadata and full description for a prefixed ID or task number.
add-task: assigns the next numeric ID. Supply a project key and type from frontmatter; the legacy defaults are shared and feature.
update-task: edits metadata, descriptions, and checklist items.
work-on-task: validates readiness and optionally starts work.
update-tasks: regenerates summary/enum tables and normalizes IDs and links.

testing means awaiting human verification, whether that is editorial review, a planning checklist, or software tests. Agents do not mark their own work done.

Optional Agent Helpers

./scripts/triage 1 --hermes
./scripts/research 1 --agy

These launch an installed Hermes or Antigravity CLI. Core task commands and tests do not need either. Select an agent with --agent, a shortcut above, or TASK_AGENT. Missing agents are reported as errors.

triage accepts planning / triage; research accepts research. Their prompts read the shared JSON configuration and refer to AGENTS.md, rather than assuming a particular project or stack. An empty commands object means no project commands are configured. Configured commands are included as guidance, not executed by these helpers.

Tooling Tests

python3 -m unittest discover -s scripts/tests -v

This validates the bundled tools, not your project's deliverables. No release scripts are provided: publishing or deployment should be added only when the project's needs are known.