agentic-jujutsu

Quantum-resistant, self-learning version control for AI agents with ReasoningBank intelligence and multi-agent coordination

1.26x

Quality

47%

Does it follow best practices?

Impact

81%

1.26x

Average score across 3 eval scenarios

Securityby

Passed

No findings from the security scan

Fix and improve this skill with Tessl

tessl review fix ./.agents/skills/agentic-jujutsu/SKILL.md

Quality

Content

50%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 rich in concrete, mostly executable API examples and covers error handling reasonably well, but it is a monolithic, marketing-flavored document roughly 2-3x longer than its instructional content justifies. Its biggest structural weakness is the absence of any progressive disclosure — API reference, use cases, and troubleshooting all belong in separate bundle files. A sequenced workflow with explicit validation checkpoints is also missing for a skill centered on concurrent, history-altering operations.

Suggestions

Split the API reference tables, 'Advanced Use Cases', and 'Examples' into separate reference files (e.g. references/API.md, references/EXAMPLES.md), keeping only Quick Start and one learning-trajectory example in SKILL.md.

Cut the marketing content — the performance comparison table, '23x faster' claims, emoji checklists, version history, and status footer — which adds ~100 lines of tokens with no instructional value.

Add a sequenced workflow with explicit validation checkpoints (install → init → record first trajectory → check getLearningStats() → act on getSuggestion() confidence), including the validation rules for finalizeTrajectory inputs as pre-flight checks rather than a standalone rules list.

Dimension	Reasoning	Score
Conciseness	The 645-line body is noticeably padded: marketing claims ('Lock-free version control (23x faster than Git)', 'Quantum-ready'), a performance comparison table, version history, a status footer, and four 'Advanced Use Cases' plus two 'Examples' that repeat the same ReasoningBank API already covered in 'Core Capabilities' and 'Best Practices'. It is not a 1 because it never lectures on concepts Claude already knows — the bloat is redundant examples and marketing, not beginner explanations.	2 / 5
Actionability	Quick Start, Core Capabilities, and the API reference tables give concrete, mostly executable code ('npx agentic-jujutsu', complete JjWrapper snippets, method signatures). It is not a 5 because several examples call undefined helpers — 'await executeOperation(op)', 'agent.analyze(diff)', an empty 'this.execute()' body, 'executeTask(agent, suggestion)' — which is pseudocode, and not a 3 because those gaps are confined to illustrative use-case sketches while the core guidance is copy-paste ready.	4 / 5
Workflow Clarity	There is no sequenced setup-to-daily-use workflow; guidance is organized by capability, not by steps. Version control is a batch/concurrent domain where validation matters, and validation appears only as scattered try/catch snippets (Use Case 3, 'Error Handling') rather than explicit checkpoints, capping this score at 3. It is not a 2 because the trajectory lifecycle (start → operate → addToTrajectory → finalize) is a coherent, consistently shown sequence with error-recovery examples.	3 / 5
Progressive Disclosure	No bundle files exist (no references/, scripts/, or assets/), so everything — full API tables, validation rules, troubleshooting, and six extended examples — is inlined in a single 645-line file; this matches 'some structure but content that should be separate is inline'. Section headers are clear and navigation is possible, and references to docs/VALIDATION_FIXES_v2.3.1.md and docs/AGENTDB_GUIDE.md are signaled but point to files not in the bundle, so it is not a 2 (headers provide real structure) nor a 4 (bulk reference content is inlined rather than split out).	3 / 5
	Total	12 / 20 Passed

Description

45%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 is a buzzword-heavy capability list with no explicit 'Use when...' trigger guidance and no concrete VCS actions. Its strongest element is naming a genuinely distinct niche (self-learning VCS for AI agents), but distinctiveness is undermined by the generic 'version control' trigger surface. Adding a trigger clause and concrete operations (commit, branch, merge, conflict resolution) would lift most dimensions.

Suggestions

Add an explicit trigger clause, e.g. 'Use when multiple AI agents need to commit, branch, and merge concurrently without locks or conflicts.'

Replace buzzwords ('Quantum-resistant', 'ReasoningBank intelligence') with concrete actions such as 'create commits, resolve conflicts automatically, and learn from past operations'.

Include natural user phrasings and variations — 'version control', 'commits', 'branches', 'merge conflicts', 'jj' — so the description matches what users actually say.

Dimension	Reasoning	Score
Specificity	The description names the domain ('version control for AI agents') but contains no concrete actions — no commit, branch, merge, or conflict-resolution verbs — while 'Quantum-resistant' and 'ReasoningBank intelligence' read as buzzwords. It sits between the vague anchor (1) and the 1-2-concrete-actions anchor (3) because the actions mentioned ('self-learning', 'multi-agent coordination') are generic marketing rather than concrete operations.	2 / 5
Completeness	A 'what' is present ('self-learning version control for AI agents') though somewhat vague, but there is no 'Use when...' clause or equivalent trigger guidance, which caps completeness at 3 per the judging guidelines. It is not a 2 because the 'what' is present rather than absent, and not a 4 because the 'when' is entirely missing rather than merely implicit.	3 / 5
Trigger Term Quality	'version control' and 'AI agents' are natural terms users would say, matching 'some relevant keywords'. It is not a 4 because common variations users actually say — commit, branch, merge, conflict, jj — are absent, and the description leans on jargon users would never utter ('ReasoningBank intelligence', 'quantum-resistant').	3 / 5
Distinctiveness Conflict Risk	'Version control' as a trigger overlaps with git/jujutsu/subversion skills, so it could fire on ordinary VCS requests; the proprietary jargon ('ReasoningBank', 'agentic') is distinctive but not something users would type. This matches 'somewhat specific but could still overlap with similar skills'.	3 / 5
	Total	11 / 20 Passed

Validation

87%

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 — 14 / 16 Passed

Validation for skill structure

Criteria	Description	Result
skill_md_line_count	SKILL.md is long (646 lines); consider splitting into references/ and linking	Warning
frontmatter_unknown_keys	Unknown frontmatter key(s) found; consider removing or moving to metadata	Warning

	Total	14 / 16 Passed

Repository: ruvnet/ruflo
Path: .agents/skills/agentic-jujutsu/SKILL.md
Commit: b14c79e

Reviewed: 3 days 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.