Blocker Minimum Input

When blocked, report exact blocker + attempts made + minimum user input needed. Enable fast unblock without back-and-forth.

Security Scan

The SKILL.md instructions align with the stated purpose (produce concise blocker reports) and the skill requests no installs, credentials, or unusual privileges — its footprint is coherent and low-risk.

Purpose & Capability

The name/description promise (report exact blocker, attempts, and minimum input) matches the instructions in SKILL.md. The file-only, instruction-only nature and absence of env vars/binaries/install spec are proportionate to the claimed functionality. Examples referencing openclaw CLI/config are reasonable for an agent-internal unblock helper.

Instruction Scope

Instructions are narrowly focused on retrying once, trying one fallback, and reporting evidence, which stays within the stated purpose. Two cautions: (1) automatic retrying of commands may have side effects if the command is destructive — the skill does not require the agent to confirm with the user before re-running potentially state-changing operations; (2) quoting error messages verbatim can accidentally expose secrets if errors include tokens — the SKILL.md says to redact tokens but does not specify how to reliably detect/redact them. These are operational precautions rather than evidence of malice.

Install Mechanism

No install spec, no code files — instruction-only reduces risk because nothing is written to disk or fetched at install time.

Credentials

The skill requests no environment variables, credentials, or config paths. Examples mention openclaw.json and CLI commands as user actions, but the skill does not ask the agent to read secrets or require platform credentials.

Persistence & Privilege

always:false and no hooks to modify other skills or system config. The skill can be invoked by the agent, which is the platform default and acceptable here.

Assessment

This skill appears coherent and low-risk: it only contains instructions for how an agent should report blockers and asks for no credentials or installs. Before enabling, consider two practical safeguards: (1) configure the agent to confirm with the user before retrying commands that might change state (to avoid unintentional destructive retries); (2) ensure the agent implements robust redaction when quoting error messages (errors can sometimes include tokens or sensitive paths). If you want stricter behavior, request the skill be modified to refuse automatic retries on potentially unsafe commands and to use a clear redaction routine for secrets.

When blocked, provide everything needed for fast unblock. No back-and-forth.

Problem

Agents often:

• Say "blocked" without specifics
• Don't report what was tried
• Ask vague questions
• Require multiple clarifications

Workflow

1. Blocker Report Format

**Blocker**: Exact error/blocker text
**Attempts Made**: 
- Attempt 1: what + result
- Attempt 2: what + result
**Minimum Unblock Input**: Smallest user action needed
**Fallback Option**: Alternative if user cannot unblock

2. Attempt Requirements

Before reporting blocker:

• Retry same approach with corrected parameters (1x)
• Try one fallback path
• If both fail, report with evidence

3. Minimum Input Examples

Blocker	Minimum Input
Missing API key	Run: openclaw configure --section web
Permission denied	Grant write access to: D:\folder OR Use fallback: C:\workspace\folder
Auth failure	Check token in: openclaw.json

Executable Completion Criteria

Criteria	Verification
Blocker text exact	Quote error message verbatim
Attempts listed	At least 2 attempts documented
Minimum input specific	Single command or action
Fallback provided	Alternative path exists

Privacy/Safety

• No credentials in blocker reports
• Redact tokens/keys from error messages
• Use generic paths when sensitive

Self-Use Trigger

Use when:

• Any tool call fails twice
• User input required to proceed
• Permission/auth issues detected

---

One report. One input. Unblock fast.