hitl

Rosetta CRITICAL MUST skill to load as Prep Step 3 Step 2 — immediately after orchestrator-contract, right before loading any workflow. Loads the session-wide approval-gate protocol governing when to stop and wait vs proceed. WITHOUT IT decisions proceed silently, violating enterprise policy. Activate it for ALL tasks always — planning, execution, validation, review — regardless of auto-mode, no approval policy, full access. THE ONLY exception: user DIRECTLY EXPLICITLY requests with EXACTLY `fully autonomous` or `No HITL`. Without explicit opt-out this skill is MANDATORY. Do not assume approval from a question or partial response. Contains human-in-the-loop collaboration, questioning, approvals, and user coordination requirements. Auto mode, full access, etc ONLY means automatic approval of tool permission prompts, HITL stays!

Quality

56%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

Securityby

Passed

No findings from the security scan

Fix and improve this skill with Tessl

tessl review fix ./instructions/r2/core/skills/hitl/SKILL.md

Quality

Content

63%Weight 40%Scale 1-5

Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.

The body is a well-structured, actionable HITL protocol with concrete rules and explicit stop/escalate/wait checkpoints. Its main weaknesses are redundancy around the approval-assumption theme and total absence of progressive disclosure — everything is inline with no reference files.

Suggestions

De-duplicate the 'do not assume approval' guidance (rules 17, 20, 27, 47, 53) into a single canonical statement to recover tokens.

Move the long HITL-gates enumeration (rules 30-42) and the full questioning checklist into a reference file referenced one level deep from SKILL.md to improve progressive disclosure.

Make the validation transitions between phases explicit (e.g., 'After approval is confirmed via the exact phrase, proceed to implementation') so the checkpoint loop is unambiguous.

Dimension	Reasoning	Score
Conciseness	The body is dense, compressed rule-style prose without concept over-explanation, but the 'do not assume approval' theme is restated across rules 17, 20, 27, 47, and 53, and several rules overlap, so it could be tightened.	3 / 5
Actionability	Concrete, specific guidance throughout: exact approval phrases ('Yes, I approve'), size buckets (SMALL/MEDIUM/LARGE), priority order (scope > security/privacy > UX > technical), and '5-10 targeted MECE questions per batch' — actionable with only minor gaps.	4 / 5
Workflow Clarity	A clear sequenced process (Questioning → Approval → HITL gates → Mismatch) with explicit checkpoints ('STOP and escalate', 'Wait for explicit user decision') and a mismatch recovery loop, though some validation transitions are implicit rather than spelled out.	4 / 5
Progressive Disclosure	Content is organized into logical sections (core_concepts, process, pitfalls) but all 59 rules live inline in one ~120-line file with no reference files to split the detail into, and nothing is signaled as one level deep.	3 / 5
	Total	14 / 20 Passed

Description

50%Weight 40%Scale 1-5

Based on the skill's description, can an agent find and select it at the right time? Clear, specific descriptions lead to better discovery.

The description conveys the skill's domain and mandatory-loading intent but is padded, directive, and lacks a clean user-facing trigger clause. Its unconditional 'all tasks always' stance hurts distinctiveness and the 'when' is not a discriminating condition.

Suggestions

Rewrite in third person as a concise capability statement ('Governs human-in-the-loop approval gates: when to stop and wait for explicit user approval versus proceed') and drop the MUST/CRITICAL/emphasis padding.

Add an explicit 'Use when...' trigger clause with natural user phrases (e.g., 'Use when the user wants approval gates, HITL review, or explicit sign-off before proceeding').

Replace 'Activate it for ALL tasks always' with discriminating conditions so the skill does not claim to overlap every task.

Dimension	Reasoning	Score
Specificity	Names the HITL/approval-gate domain and a few concrete actions ('questioning, approvals, and user coordination requirements', 'governing when to stop and wait vs proceed'), but the capability list is thin and buried under heavy MUST/CRITICAL padding about loading order rather than what the skill does.	3 / 5
Completeness	A 'what' is present ('Loads the session-wide approval-gate protocol...') and a 'when' exists ('Activate it for ALL tasks always' with an opt-out exception), but there is no 'Use when...' clause and the 'when' is unconditional rather than a discriminating trigger, capping completeness at 3 per the guideline.	3 / 5
Trigger Term Quality	Relevant terms appear ('human-in-the-loop', 'HITL', 'approvals', 'approval-gate', 'questioning') but they are embedded in directive language aimed at Claude rather than natural user phrases, and common variations are missing.	3 / 5
Distinctiveness Conflict Risk	The HITL/approval-gate niche is itself distinct, but the explicit 'Activate it for ALL tasks always — planning, execution, validation, review' framing creates high overlap risk with virtually every other skill.	3 / 5
	Total	12 / 20 Passed

Validation

93%

Warnings & errors only

Checks the skill against the spec for correct structure and formatting. All validation checks must pass before discovery and implementation can be scored.

Validation — 15 / 16 Passed

Validation for skill structure

Criteria	Description	Result
frontmatter_unknown_keys	Unknown frontmatter key(s) found; consider removing or moving to metadata	Warning

	Total	15 / 16 Passed

Repository: griddynamics/rosetta
Path: instructions/r2/core/skills/hitl/SKILL.md
Commit: 044d86c

Reviewed: 20 hours ago

Table of Contents

Discovery Implementation Validation

Is this your skill?

If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.