Grill
Interrogate consequential uncertainty when a user explicitly asks for rigorous clarification or a non-trivial task has unresolved goals, constraints, tradeoffs, or acceptance signals. Skip settled work and low-cost micro-decisions.
Install
$ agentstack add skill-aquish-lee-agent-handrails-grill ✓ scanned · ✓ verified, works with Claude Code, Cursor, and more.
Security review
✓ PassedNo issues found. Passed automated security review. · v0.1.0 How review works →
- ✓ Prompt-injection patterns
- ✓ Secret / credential exfiltration
- ✓ Dangerous shell & filesystem operations
- ✓ Untrusted network calls
- ✓ Known-malicious package signatures
What it can access
- ✓ Network access No
- ✓ Filesystem access No
- ✓ Shell / process execution No
- ✓ Environment & secrets No
- ✓ Dynamic code execution No
From automated source analysis of v0.1.0. “Used” means the capability is present in the source — more access means more to trust, not that it’s unsafe.
Verified badge
Passed review? Show it. Paste this badge into your README, it links to the public security report.
Links to your public security report.[](https://agentstack.voostack.com/security/report/skill-aquish-lee-agent-handrails-grill) Reliability & compatibility
Declared compatibility
Compatibility is declared by the source manifest. End-to-end runtime verification is coming, see below.
We're building live execution health for every listing: tool-call success rate, median latency, uptime, and last-checked timestamps, measured, not self-reported. It isn't live yet, so we don't show numbers we can't stand behind.
How agent discovery & health will work →About
Grill
Resolve only the uncertainty that could materially change the outcome, architecture, authorization, or verification strategy.
Read [the local v2 contract](references/usable-loop-v2.md) for the full Alignment Snapshot and Outcome Reference shapes.
Calibrate the questioning
Do not question for ceremony. Skip or stop when the task is already decision-complete. Focus on uncertainties with high consequence and low reversibility:
- what result must be true for the user;
- who or what has authority to define that result;
- constraints, non-goals, compatibility, and safety boundaries;
- the tradeoff between plausible approaches;
- observable acceptance and failure signals.
Ask one compact cluster at a time. Explain why an answer changes the design when that is not obvious. Use repository evidence for discoverable facts instead of asking the user to repeat them.
Authority handling
A current request is authoritative when the user owns its scope and no source or permission conflict exists. Agent-created options and interpretations remain provisional until selected. Do not let a saved spec, plan, Capsule, or previous answer silently override the current designated authority.
Converge
When the consequential questions are settled, return a compact Alignment Snapshot:
- Outcome Reference;
- task shape and goal;
- non-goals;
- chosen direction and rejected consequential alternatives;
- key constraints and decisions;
- open risks or unknowns;
- verification signals;
- an optional suggested next handrail.
A suggestion is not an invocation. Do not create a spec or plan unless the user asked for one. For a micro-task, a short conversational alignment is enough.
Direct result
Lead with the resolved decision, then remaining material uncertainty and its impact. Expand a full Checkpoint only for a continuity trigger.
Source & license
This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.
- Author: Aquish-Lee
- Source: Aquish-Lee/agent-handrails
- License: MIT
Install and usage instructions live in the source repository linked above.
Reviews
No reviews yet, be the first.
Write a review
Versions
- v0.1.0 Imported from the upstream source.