Skill Budget

Direct Invocation

• User explicitly runs /budget command with task number
• User runs /research on a task with language "present" and task_type "budget"

Implicit Invocation (during task implementation)

When an implementing agent encounters any of these patterns:

Plan step language patterns:

• "Create grant budget spreadsheet"
• "Build budget with F&A calculation"
• "Generate NIH/NSF budget"
• "Grant budget with salary cap"

Target mentions:

• "grant budget"
• "NIH budget"
• "NSF budget"
• "F&A calculation"
• "indirect cost budget"

When NOT to trigger

Do not invoke for:

• Budget narrative/justification text (use skill-grant with --budget)
• General cost breakdowns (use skill-spreadsheet from founder)
• Market sizing or revenue projections

Execution Flow

Stage 1: Input Validation

Validate required inputs:

• task_number - Must be provided and exist in state.json
• mode - Optional, one of: MODULAR, DETAILED, NSF, FOUNDATION, SBIR

# Lookup task
task_data=$(jq -r --argjson num "$task_number" \
  '.active_projects[] | select(.project_number == $num)' \
  specs/state.json)

# Validate exists
if [ -z "$task_data" ]; then
  return error "Task $task_number not found"
fi

# Extract fields
task_type=$(echo "$task_data" | jq -r '.task_type // "present"')
status=$(echo "$task_data" | jq -r '.status')
project_name=$(echo "$task_data" | jq -r '.project_name')
description=$(echo "$task_data" | jq -r '.description // ""')

# Extract pre-gathered forcing_data (if present)
forcing_data=$(echo "$task_data" | jq -r '.forcing_data // null')
pre_gathered_mode=$(echo "$forcing_data" | jq -r '.mode // null' 2>/dev/null)

# Validate mode if provided
if [ -n "$mode" ]; then
  case "$mode" in
    MODULAR|DETAILED|NSF|FOUNDATION|SBIR) ;;
    *) return error "Invalid mode: $mode. Must be MODULAR, DETAILED, NSF, FOUNDATION, or SBIR" ;;
  esac
fi

Stage 2: Preflight Status Update

Update task status to "researching" BEFORE invoking subagent.

Update state.json:

jq --arg ts "$(date -u +%Y-%m-%dT%H:%M:%SZ)" \
   --arg status "researching" \
   --arg sid "$session_id" \
  '(.active_projects[] | select(.project_number == '$task_number')) |= . + {
    status: $status,
    last_updated: $ts,
    session_id: $sid
  }' specs/state.json > specs/tmp/state.json && mv specs/tmp/state.json specs/state.json

Update TODO.md: Use Edit tool to change status marker to [RESEARCHING].

Stage 3: Create Postflight Marker

padded_num=$(printf "%03d" "$task_number")
mkdir -p "specs/${padded_num}_${project_name}"

cat > "specs/${padded_num}_${project_name}/.postflight-pending" << EOF
{
  "session_id": "${session_id}",
  "skill": "skill-budget",
  "task_number": ${task_number},
  "operation": "research",
  "reason": "Postflight pending: status update, artifact linking, git commit",
  "created": "$(date -u +%Y-%m-%dT%H:%M:%SZ)"
}
EOF

Stage 4: Prepare Delegation Context

Include pre-gathered forcing_data when available:

{
  "task_context": {
    "task_number": N,
    "project_name": "{project_name}",
    "description": "{description}",
    "task_type": "present",
    "task_type": "budget"
  },
  "forcing_data": {
    "mode": "{pre_gathered_mode}",
    "project_period": "{pre_gathered_period}",
    "direct_cost_cap": "{pre_gathered_cap}",
    "gathered_at": "{timestamp}"
  },
  "mode": "MODULAR|DETAILED|NSF|FOUNDATION|SBIR or use forcing_data.mode",
  "metadata_file_path": "specs/{NNN}_{SLUG}/.return-meta.json",
  "metadata": {
    "session_id": "sess_{timestamp}_{random}",
    "delegation_depth": 1,
    "delegation_path": ["orchestrator", "budget", "skill-budget"]
  }
}

Note: If forcing_data is present from STAGE 0 of /budget command, pass it to the agent. The agent will use pre-gathered data and only ask follow-up questions for missing details.

Stage 5: Invoke Agent

CRITICAL: You MUST use the Task tool to spawn the agent.

Required Tool Invocation:

Tool: Task (NOT Skill)
Parameters:
  - subagent_type: "budget-agent"
  - prompt: [Include task_context, forcing_data, mode, metadata_file_path, metadata]
  - description: "Grant budget spreadsheet generation"

The agent will:

• Use pre-gathered forcing_data if available (skip already-answered questions)
• Present mode selection only if not pre-selected
• Ask follow-up forcing questions for budget details
• Generate multi-year XLSX with native Excel formulas
• Export JSON metrics
• Create research report at specs/{NNN}_{SLUG}/reports/
• Write metadata file
• Return brief text summary

Stage 5b: Self-Execution Fallback

CRITICAL: If you performed the work above WITHOUT using the Task tool (i.e., you read files, wrote artifacts, or updated metadata directly instead of spawning a subagent), you MUST write a .return-meta.json file now before proceeding to postflight. Use the schema from return-metadata-file.md with the appropriate status value for this operation.

If you DID use the Task tool, skip this stage -- the subagent already wrote the metadata.

Postflight (ALWAYS EXECUTE)

The following stages MUST execute after work is complete, whether the work was done by a subagent or inline (Stage 5b). Do NOT skip these stages for any reason.

Stage 6: Parse Subagent Return

padded_num=$(printf "%03d" "$task_number")
metadata_file="specs/${padded_num}_${project_name}/.return-meta.json"

if [ -f "$metadata_file" ] && jq empty "$metadata_file" 2>/dev/null; then
    status=$(jq -r '.status' "$metadata_file")
    artifacts=$(jq -r '.artifacts' "$metadata_file")
else
    status="failed"
fi

Stage 7: Update Task Status (Postflight)

If status is "researched", update state.json and TODO.md.

Update state.json:

jq --arg ts "$(date -u +%Y-%m-%dT%H:%M:%SZ)" \
   --arg status "researched" \
  '(.active_projects[] | select(.project_number == '$task_number')) |= . + {
    status: $status,
    last_updated: $ts
  }' specs/state.json > specs/tmp/state.json && mv specs/tmp/state.json specs/state.json

Update TODO.md: Use Edit tool to change status marker to [RESEARCHED].

Stage 8: Link Artifacts

Add artifacts to state.json with summaries.

IMPORTANT: Use two-step jq pattern to avoid escaping issues.

# For each artifact type (research, spreadsheet, metrics)
for artifact in $(echo "$artifacts" | jq -c '.[]'); do
    path=$(echo "$artifact" | jq -r '.path')
    type=$(echo "$artifact" | jq -r '.type')
    summary=$(echo "$artifact" | jq -r '.summary')

    # Step 1: Filter out existing artifacts of same type (use "| not" pattern)
    jq '(.active_projects[] | select(.project_number == '$task_number')).artifacts =
        [(.active_projects[] | select(.project_number == '$task_number')).artifacts // [] | .[] | select(.type == "'$type'" | not)]' \
      specs/state.json > specs/tmp/state.json && mv specs/tmp/state.json specs/state.json

    # Step 2: Add new artifact
    jq --arg path "$path" \
       --arg type "$type" \
       --arg summary "$summary" \
      '(.active_projects[] | select(.project_number == '$task_number')).artifacts += [{"path": $path, "type": $type, "summary": $summary}]' \
      specs/state.json > specs/tmp/state.json && mv specs/tmp/state.json specs/state.json
done

Update TODO.md: Link artifact using count-aware format.

Apply the four-case Edit logic from @.claude/context/patterns/artifact-linking-todo.md with field_name=**Research**, next_field=**Plan**.

Stage 9: Git Commit

git add -A
git commit -m "task ${task_number}: complete research

Session: ${session_id}

Stage 10: Cleanup

rm -f "specs/${padded_num}_${project_name}/.postflight-pending"
rm -f "specs/${padded_num}_${project_name}/.postflight-loop-guard"
rm -f "specs/${padded_num}_${project_name}/.return-meta.json"

Stage 11: Return Brief Summary

Grant budget research completed for task {N}:
- Mode: {mode}, {questions_asked} forcing questions completed
- Funder: {funder_type}, {num_years}-year project
- Personnel: {count} ({roles})
- Year 1 direct costs: ${year1_direct}
- Total project cost: ${total}
- Pre-gathered data used: {yes/no}
- Spreadsheet: specs/{NNN}_{SLUG}/grant-budget.xlsx
- Research report: specs/{NNN}_{SLUG}/reports/{NN}_{short-slug}.md
- Status updated to [RESEARCHED]
- Changes committed
- Next: Run /plan {N} to create implementation plan

Return Format

Brief text summary (NOT JSON).

Error Handling

Input Validation Errors

Return immediately if task not found.

Metadata File Missing

Keep status as "researching" for resume.

User Abandonment

Return partial status with progress made.

Git Commit Failure

Non-blocking: Log failure but continue.