CtrlK
BlogDocsLog inGet started
Tessl Logo
SkillPublic

jarvis

Load when acting for the user on their computer, browser, accounts, messages or records, or when starting, handing off or ending a long session.

Quality

64%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide
SecuritybySnyk

Low

Low-risk findings worth noting

Fix and improve this skill with Tessl

tessl review fix ./.agents/skills/jarvis/SKILL.md
SKILL.md
Quality
Evals
Security

Quality

Content

57%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 lean, well-structured orchestrator overview with a genuinely executable reachability probe, but as bundled it is not self-sufficient: the install, persona and workflow routing steps point at files that are absent, and the destructive cleanup guidance lacks explicit validation checkpoints. Shipping the missing workflows/, persona.md and README (or removing those routes) plus a verification step before removal would move this to good.

Suggestions

Include the missing workflows/ files (session, docs-audit, efficiency-audit, the four setup guides) plus persona.md and README.md in the bundle, or drop their routes from the reading-routes table so no link is dead.

Add an explicit validation checkpoint to the Cleanup section — e.g. enumerate what will be removed, confirm with the user or a check command, then remove and re-verify — so the destructive workflow has a verify→fix→retry loop.

Reference references/control-host-events.md and references/long-session.md from the reading-routes table (or move them into the files that use them) so no bundled reference is undiscoverable from the entry file.

DimensionReasoningScore

Conciseness

The ~65-line body is lean and assumes competence: no concept explanations, no padding, dense imperative instructions like "Back up only if a step can lose the only copy; delete the backup after checking the result." It falls short of anchor 5 only through minor redundancy (the agents/install paragraph restates "Copying only the skill folder… supplies none of their harness agent files" then "Installing `orch` alone does not install Jarvis") and a few over-compressed sentences (the "judge each folder `kendex project list` prints, a `(worktree of [MAIN])` line" phrasing) that could be tightened for clarity.

4 / 5

Actionability

Some guidance is genuinely executable — the exact mailbox probe command ("lane-mail events --item overseer --box to-overseer") with exit-code interpretation, "kendex refresh", and the one-line persona offer procedure — but key steps point to artifacts absent from the bundle: the install step routes to README.md#install, the persona steps depend on persona.md, and the routing table sends 7 tasks to workflows/*.md files, none of which exist here. That matches anchor 3 (some concrete guidance but incomplete, missing key details) rather than 4.

3 / 5

Workflow Clarity

The reachable-repositories procedure is the strongest part: a concrete probe, explicit branch interpretation (exit 0 reachable; anything else unreachable with its cause; orch-without-mailbox → "refresh needed" offer), and a defined reporting step. However, the destructive "Cleanup and backups" section ("Remove all unused code, files, keys, secret-store entries, databases, records and settings from every holder in the same change") has no explicit validation or verification loop before removal — validation is only implicit ("after checking the result") or deferred to a reference — so per the destructive-operation cap workflow clarity cannot exceed 3.

3 / 5

Progressive Disclosure

The structure is a good overview-plus-routing pattern: a task-to-file table with clearly signaled one-level-deep references, and 11 of the referenced references/*.md files exist. But scored against the actual bundle, 8 of the in-bundle targets are missing (README.md, persona.md, and all seven workflows/*.md), and references/ holds two files (control-host-events.md, long-session.md) the body never points to. Navigation is better organized than anchor 3's "references present but not clearly signaled" yet the broken links are more than the "minor organization gaps" of anchor 4.

3 / 5

Total

13

/

20

Passed

Description

57%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 pure trigger clause with excellent "when" coverage and natural keywords, but it never states what the skill does — the capability statement was relegated to the summary field. Moving a compact capability list into the description would lift completeness and specificity substantially.

Suggestions

Prefix the description with what the skill does, e.g. "Guides a delegated desktop assistant through user decisions, checked results and communication. Load when…", so both what and when appear in the description field.

Add common trigger synonyms such as "email", "sign-in/logins", "passwords/credentials" and "handover" to the when-clause to widen natural matching.

Trim one or two of the broadest surfaces (e.g. "records") that overlap other skills' triggers, or qualify them (e.g. "your accounts and records") to reduce conflict risk.

DimensionReasoningScore

Specificity

The description names several concrete domains ("computer, browser, accounts, messages or records") and concrete session-lifecycle operations ("starting, handing off or ending a long session"), but the core action remains the generic "acting for the user". It matches anchor 3 (domain plus a couple of actions, not comprehensive) rather than 2, since the session-lifecycle clause names real operations; it falls short of 4 because no distinct capability of the skill itself is enumerated.

3 / 5

Completeness

The "when" is explicit and multi-faceted ("Load when acting for the user… or when starting, handing off or ending a long session"), but the "what" — what the skill actually does — is only vaguely implied by "acting for the user"; the capability statement lives in the summary field, not the description. Anchor 2 ("only 'when' present without 'what'") nearly fits, but the delegation phrasing weakly implies the what, so this sits at 3; it cannot reach 4 because the what is never stated.

3 / 5

Trigger Term Quality

Natural user-facing keywords are well covered: "computer", "browser", "accounts", "messages", "records", "session", "handing off" — phrases a user would plausibly say when delegating. It misses common synonyms and variations (e.g., email, login, passwords, handover), which keeps it below anchor 5 but comfortably above anchor 3.

4 / 5

Distinctiveness Conflict Risk

The delegated-assistant framing plus session-handoff triggers carve a recognizable niche, but the enumerated surfaces ("computer, browser, accounts, messages or records") span the territory of browser-automation, communication and session-management skills, creating real overlap risk. It is more specific than anchor 2's "very broad" but not the "minor overlap risk" of anchor 4.

3 / 5

Total

13

/

20

Passed

Validation

81%

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

Validation — 13 / 16 Passed

Validation for skill structure

CriteriaDescriptionResult

metadata_version

'metadata.version' is missing

Warning

frontmatter_unknown_keys

Unknown frontmatter key(s) found; consider removing or moving to metadata

Warning

relative_links

Relative link issues: 9 missing, 1 suspicious

Warning

Total

13

/

16

Passed

Repository

vanillagreencom/kendex

Path

.agents/skills/jarvis/SKILL.md

Reviewed

Table of Contents

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.