Agent Step Sequencer
v1.1.0Multi-step scheduler for in-depth agent requests. Detects when user needs multiple steps, suggests plan and waits for confirmation, persists state, and runs heartbeat-aware flow. Use when requests have 3+ actions, sequential dependencies, output dependencies, or high scope/risk.
⭐ 3· 2.1k·5 current·7 all-time
bygostlight@gostlightai
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
OpenClaw
Suspicious
medium confidencePurpose & Capability
The stated purpose (multi-step, heartbeat-aware scheduling and recovery) matches the shipped artifacts: runner/check Python scripts, state schema, and step/heartbeat semantics. However the registry metadata claims no required binaries/env and 'no install spec', while the embedded SKILL.md front-matter declares python3 and an env var STEP_AGENT_CMD and includes apt/brew install hints — that discrepancy is unexpected and should be reconciled.
Instruction Scope
Runtime instructions tell the agent to persist state.json, read/write workspace-relative paths, invoke scripts, and autonomously retry/try alternative data sources without asking the user mid-step. Those behaviors are consistent with the skill but expand agent autonomy (file writes, external API calls, retries and alternatives). If you require explicit human approval between steps or tighter limits on outbound calls, this skill's instructions grant broad autonomous discretion.
Install Mechanism
No install spec was reported in the registry, yet SKILL.md includes apt/brew install metadata for python3 and the package bundle contains runnable Python scripts. There are no high-risk remote downloads or archive extracts; the code is local and uses standard Python. The install story is inconsistent (instruction-only vs bundled scripts + install hints), which is a packaging/manifest issue to fix.
Credentials
The SKILL.md declares a single env var STEP_AGENT_CMD (how the runner invokes the agent) and a dependency on python3 — both are proportionate to a runner/checker. The registry-level summary omitted these requirements, creating an incoherence. The skill requires filesystem write/read access to persist state.json and check requiredOutputs; no secret credentials are requested.
Persistence & Privilege
The skill persists state to workspace state.json, supports heartbeat-driven autonomous resumption, and allows the agent to execute steps without asking the user mid-step. Flags do not set always:true, but model invocation is not disabled. This combination enables long-lived autonomous behavior and workspace mutation — appropriate for the stated purpose but deserving user scrutiny and explicit policy controls.
What to consider before installing
This skill appears to implement a legitimate multi-step scheduler, but there are inconsistencies in the package metadata vs the SKILL.md. Before installing: (1) inspect scripts/step-sequencer-runner.py and step-sequencer-check.py to verify they do only what you expect (look for network calls, exec of arbitrary commands, credential access). (2) Ensure you are comfortable with the skill writing state.json in the workspace and the agent performing autonomous retries/alternative-source actions. (3) Provide a safe STEP_AGENT_CMD and verify python3 availability, or update the manifest to match the SKILL.md. (4) If you want manual approval between steps, either modify the skill or set policy to require user confirmation (e.g., disable automatic model invocation). If you cannot review the code, treat the packaging inconsistencies as a red flag and avoid installing until they are resolved.Like a lobster shell, security has layers — review code before you run it.
latestvk97f001awgyvmncez4bte2qr1h80xrt3
License
MIT-0
Free to use, modify, and redistribute. No attribution required.
Runtime requirements
🔗 Clawdis
Binspython3
EnvSTEP_AGENT_CMD
Install
Install Python 3
Bins: python3
brew install python@3