Subagent Spawn Command Builder

Build sessions_spawn command payloads from JSON profiles. Use when you want reusable subagent profiles (model/thinking/timeout/cleanup/agentId/label) and com...

Security Scan

The skill does what it claims—build JSON payloads for sessions_spawn and log them locally—its files, instructions, and required resources are consistent with that purpose.

Purpose & Capability

Name/description match the included code and SKILL.md: the script builds a sessions_spawn payload JSON from CLI args and does not attempt to execute any spawn operation or call external services.

Instruction Scope

SKILL.md and the script are narrowly scoped to parsing CLI args, constructing a payload object, writing a JSON log line to state/build-log.jsonl, and printing the payload to stdout. The instructions do not read unrelated system files or environment variables, nor do they send data to external endpoints.

Install Mechanism

No install spec is provided (instruction-only with a small included Node script). There are no downloads or package installs; risk from installation is minimal.

Credentials

The skill requests no environment variables, credentials, or config paths. All inputs come from CLI arguments, which is proportional to its stated purpose.

Persistence & Privilege

The script creates a local state directory and appends log lines to state/build-log.jsonl containing timestamp, profile label, and the generated payload (including the full task text and any provided fields). This is expected for a builder but means user-supplied text will be persisted on disk under the skill's state directory.

Assessment

This skill appears coherent and limited in scope: it only builds JSON payloads and logs them locally. Before installing, be aware that every run will append a JSON line to state/build-log.jsonl containing the full payload (including the --task text and any other flags you pass). If your tasks contain sensitive data or secrets, avoid passing them to this builder or ensure the state log is stored securely/cleaned up. Also note the script does not execute subagents or make network calls. If you rely on default profile values referenced in TOOLS.md, verify that TOOLS.md exists in your environment and provides the values you expect.

Generate sessions_spawn payload JSON from profile config. This skill does not execute sessions_spawn; it only builds payload/command JSON.

Files

• Builder script: scripts/build_spawn_payload.mjs
• Builder log: state/build-log.jsonl

Supported sessions_spawn parameters

• task (required)
• label (optional)
• agentId (optional)
• model (optional)
• thinking (optional)
• runTimeoutSeconds (optional)
• cleanup (keep|delete, optional)
• cwd (optional) — working directory for the subagent
• mode (run|session, optional)

Setup

Read the "Subagent Spawn Profiles" table in TOOLS.md for default values per profile. Pass values as explicit CLI arguments (--model, --thinking, --run-timeout-seconds, --cleanup). The --profile flag is now a logging label, not a lookup key for a config file.

Generate payload

skills/subagent-spawn-command-builder/scripts/build_spawn_payload.mjs \
  --profile heartbeat \
  --task "Analyze recent context and return a compact summary" \
  --label heartbeat-test \
  --model claude-sonnet-4-20250514 \
  --thinking low \
  --run-timeout-seconds 300 \
  --cleanup delete

The script prints JSON directly usable for sessions_spawn.

Merge/priority rule

All values come from explicit CLI arguments. --profile is a logging label only (not a config lookup key). Refer to the "Subagent Spawn Profiles" table in TOOLS.md for recommended defaults per profile.

task always comes from CLI --task.

CLI options

Note: this builder is Node.js (.mjs) based. If generated tasks include Python execution steps, write commands with python3 (not python).

• --profile (required)
• --task (required)
• --label
• --agent-id
• --model
• --thinking
• --run-timeout-seconds
• --cleanup keep|delete
• --cwd <path>
• --mode run|session