Name

ai-dlc:resume - Resume an existing AI-DLC intent.

Synopsis

code

/resume [intent-slug]

Description

User-facing command - Resume work on an intent when ephemeral state (iteration.json) is lost but .ai-dlc/ artifacts exist.

This happens when:

•Session context was cleared unexpectedly
•Starting fresh session after previous work
•Branch state lost but artifacts preserved

User Flow:

code

SessionStart: "Resumable Intents Found: my-feature"
User: /resume my-feature
AI: Initialized state, continuing as builder...
User: /construct
...continues work...

Implementation

Step 1: Find Resumable Intents

If no slug provided, scan multiple sources for active intents:

A: Check filesystem first (highest priority - source of truth):

bash

for intent_file in .ai-dlc/*/intent.md; do
  [ -f "$intent_file" ] || continue
  dir=$(dirname "$intent_file")
  slug=$(basename "$dir")
  status=$(han parse yaml status -r --default active < "$intent_file")
  [ "$status" = "active" ] && echo "$slug"
done

B: Check git branches using discover_branch_intents:

bash

# Source DAG library
source "${CLAUDE_PLUGIN_ROOT}/lib/dag.sh"

# Discover from branches (include remote for resume command)
branch_intents=$(discover_branch_intents true)

# Parse results: slug|workflow|source|branch
echo "$branch_intents" | while IFS='|' read -r slug workflow source branch; do
  [ -z "$slug" ] && continue
  echo "$slug ($source: $branch)"
done

Selection logic:

•1 intent found -> auto-select
•Multiple intents -> list them and prompt user to specify
•0 intents -> error, suggest /elaborate

Step 2: Load Intent Metadata

Read from .ai-dlc/{slug}/intent.md:

javascript

const intentFile = `.ai-dlc/${slug}/intent.md`;
const workflow = han_parse_yaml(intentFile, "workflow") || "default";
const title = han_parse_yaml(intentFile, "title") || slug;

Step 3: Determine Starting Hat

Use DAG analysis to determine where to resume:

bash

# Source DAG library
source "${CLAUDE_PLUGIN_ROOT}/lib/dag.sh"

# Get recommended hat based on unit states
starting_hat=$(get_recommended_hat ".ai-dlc/${slug}" "${workflow}")

Hat selection logic:

•No units exist -> planner (needs decomposition)
•All units completed -> reviewer (final verification)
•Any units in_progress or ready -> builder (continue work)
•All units blocked -> planner (resolve dependencies)

Step 4: Create Intent Worktree

CRITICAL: The orchestrator MUST run in the intent worktree, not the main working directory.

bash

INTENT_BRANCH="ai-dlc/${slug}"
INTENT_WORKTREE="/tmp/ai-dlc-${slug}"

if [ ! -d "$INTENT_WORKTREE" ]; then
  git worktree add -B "$INTENT_BRANCH" "$INTENT_WORKTREE"
fi
cd "$INTENT_WORKTREE"

Step 5: Initialize State

Save to han keep storage (intent-level state goes to current branch, which is now the intent branch):

bash

# Save intent slug (intent-level state -> current branch / intent branch)
han keep save intent-slug "$SLUG"

# Save iteration state (intent-level state -> current branch / intent branch)
han keep save iteration.json "{\"iteration\":1,\"hat\":\"$STARTING_HAT\",\"workflowName\":\"$WORKFLOW\",\"workflow\":$WORKFLOW_HATS_JSON,\"status\":\"active\"}"

Step 5b: Restore Team (Agent Teams)

When CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS is enabled:

bash

AGENT_TEAMS_ENABLED="${CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS:-}"
TEAM_NAME="ai-dlc-${SLUG}"
TEAM_CONFIG="$HOME/.claude/teams/${TEAM_NAME}/config.json"

If AGENT_TEAMS_ENABLED is set:

•
Team exists (TEAM_CONFIG found):
- •Read config to discover active teammates
- •Save teamName to iteration.json
- •Note: Active teammates may need to be re-spawned if they were shut down
•
Team does not exist:
- •Save teamName to iteration.json (for /construct to create it)
- •Note in output: "Team will be created when /construct runs"

Without Agent Teams: Skip this step. No team management needed.

Step 6: Output Confirmation

markdown

## AI-DLC Intent Resumed

**Intent:** {title}
**Slug:** {slug}
**Workflow:** {workflow}
**Starting Hat:** {startingHat}
**Worktree:** /tmp/ai-dlc-{slug}/
**Team:** {teamName} (if Agent Teams enabled)

### Unit Status
{get_dag_status_table output}

**Summary:** {completed}/{total} units completed

**Next:** Run `/construct` to continue the construction loop.

Note: All AI-DLC work happens in the worktree at /tmp/ai-dlc-{slug}/

Examples

Single Intent (Auto-Select)

code

User: /resume
AI: Found 1 resumable intent: han-team-platform

## AI-DLC Intent Resumed

**Intent:** Han Team Platform
**Slug:** han-team-platform
**Workflow:** default
**Starting Hat:** builder
**Worktree:** /tmp/ai-dlc-han-team-platform/

### Unit Status
| Unit | Status | Blocked By |
|------|--------|------------|
| unit-01-core-setup | completed | |
| unit-02-auth | in_progress | |
| unit-03-dashboard | pending | unit-02-auth |

**Summary:** 1/3 units completed

**Next:** Run `/construct` to continue the construction loop.

Multiple Intents (Requires Selection)

code

User: /resume
AI: Found multiple resumable intents:
- han-team-platform (default workflow, 1/3 completed)
- api-refactor (tdd workflow, 0/5 completed)

Please specify: `/resume han-team-platform` or `/resume api-refactor`