td-task-management

td - Task Management for AI Agents

Overview

td is a minimalist CLI for tracking tasks and maintaining agent memory across context windows. When your AI session ends, td captures what was done, what remains, and what decisions were made—so the next session picks up exactly where the last one left off.

Core capability: Run td usage and get everything needed for the next action—current focus, pending reviews, open issues, recent decisions.

Quick Start

Starting a New Agent Context

td usage --new-session -q  # Auto-rotate and show compact current state

Use td usage without -q when workflow guidance would be useful. Status output includes:

Active work sessions and recent decisions
What issues are pending review (you can review these)
Highest priority open issues
Recent handoffs from previous sessions

Single-Issue Workflow

For focused work on one issue:

td start <issue-id>                    # Begin work
td log "OAuth callback implemented"    # Track progress
td log --decision "Using JWT tokens"   # Log decisions
td handoff <id> --done "..." --remaining "..."  # Capture state
td review <id>                         # Submit for review

Multi-Issue Workflow

For agents handling related issues:

td ws start "Auth implementation"      # Start work session
td ws tag td-a1b2 td-c3d4             # Associate issues (auto-starts them)
td ws tag --no-start td-e5f6          # Associate without starting
td ws log "Shared token storage"       # Log to all tagged issues
td ws handoff                          # Capture state, end session

Key Workflows

Workflow 1: Starting New Work

# 1. Check what to work on
td usage          # See current state
td next           # Highest priority open issue
td critical-path  # What unblocks most work

# 2. Start work
td start <id>

# 3. Begin logging
td log "Started implementation"

Workflow 2: Handing Off Work

Use a structured handoff when work will continue in another context:

td handoff <id> \
  --done "OAuth flow, token storage" \
  --remaining "Refresh token rotation, error handling" \
  --decision "Using JWT for stateless auth" \
  --uncertain "Should tokens expire on password change?"

Keys:

--done - What is complete
--remaining - What remains
--decision - Why you chose approach X
--uncertain - What you're unsure about

Next session will see all this context with td usage or td context <id>.

Workflow 3: Reviewing Code

# 1. See reviewable issues
td reviewable

# 2. Check details
td show <id>
td context <id>

# 3. Approve or reject
# Independent review:
td approve <id> --reason "Reviewed diff, looks good"
# Or, in trusted mode, when you implemented it yourself:
#   a sub-agent reviewed it — name them (no --reason needed):
td approve <id> --reviewed-by "code-reviewer sub-agent"
#   you reviewed it yourself — say so:
td approve <id> --self-review --reason "Reviewed own diff, tests pass"
# Or:
td reject <id> --reason "Missing error handling"

Independent review is preferred when practical. The default trusted mode also lets an involved session approve by saying who reviewed the work: --reviewed-by "<who>" credits someone else, --self-review --reason owns it. Both are recorded and neither is verifiable by td, so never name a reviewer who did not review — that reads as independent in the audit trail, which is worse than an honest self-review. In delegated and strict an involved session cannot approve at all.

Workflow 4: Handling Blockers

# 1. Log the blocker
td log --blocker "Waiting on API spec from backend team"

# 2. Work on something else
td next              # Get another issue
td ws tag td-e5f6   # Add to work session

# 3. Come back to blocked issue later
td context td-a1b2  # Refresh context when blocker resolves

Commands by Category

Checking Status

td usage - Current state, reviews, next steps
td usage -q - Compact current state
td current - What you're working on
td ws current - Current work session state
td next - Highest priority open
td critical-path - What unblocks most work

Working on Issues

td start <id> - Begin work
td unstart <id> - Revert to open (undo accidental start)
td log "msg" - Track progress
td log --decision "..." - Log decision
td log --blocker "..." - Log blocker
td show <id> - View details
td context <id> - Full context for resuming

Handing Off

td handoff <id> --done "..." --remaining "..." - Single issue
td ws handoff - Multi-issue work session

Reviews

td review <id> - Submit for review
td reviewable - Issues you can review
td approve <id> --reason "..." - Approve an independent review
td approve <id> --reviewed-by "<who>" - Record who reviewed it when that is not you (trusted mode)
td approve <id> --self-review --reason "..." - Acknowledge and record a trusted-mode self-review
td approve <id> --record-only --reason "..." - Attest without closing; any session closes after
td reject <id> --reason "..." - Reject

Creating/Managing Issues

td create "title" --type feature --priority P1 - Create
td create "title" --description-file body.md --acceptance-file acceptance.md - Agent-safe rich text
cat body.md | td update <id> --append --description-file - - Append rich text from stdin
td list - List all
td list --status in_progress - Filter by status
td block <id> - Mark as blocked
td delete <id> - Delete

File Tracking

td link <id> <files...> - Track files with issue
td files <id> - Show file changes

Other

td monitor - Live dashboard
td session --new "name" - Force new session
td undo - Undo last action

See quick_reference.md for full command listing.

Resources

quick_reference.md

Complete command reference organized by task type.

ai_agent_workflows.md

Detailed workflows for common AI agent scenarios:

Single-issue focus
Multi-issue work sessions
Handling blockers
Resuming work
Code review process
Tips for AI agents

Issue Lifecycle

open → in_progress → in_review → closed
         |              |
         v              | (reject)
     blocked -----------+

Key Principles

Session Isolation: Every terminal/context gets a session ID for continuity and audit. Independent review is preferred; trusted mode also lets an involved session approve by recording who actually reviewed the work.

Structured Handoffs: Record done/remaining/decisions/uncertain when another context will need to continue the work.

Minimal: Does one thing. Single binary, SQLite local storage (.todos/), no server, works with any AI tool.

For AI Agents

At the start of a new agent context:

td usage --new-session -q

This auto-rotates sessions and gives you compact current state. Use judgment about how much tracking detail each task needs:

Single focused issue → Use single-issue workflow
Multiple related issues → Use td ws start for work sessions
Work will continue elsewhere → Use td handoff or td ws handoff
A decision aids continuity → Log it with --decision
An uncertainty matters → Record it with --uncertain
Track files → Use td link so future sessions know what changed

See ai_agent_workflows.md for detailed examples.