CategoryWeightCategory scoreContribution

Securityprompt, exec, net, exfil, eval

35%

26.6 pts

Supply chainhash, typosquat, maintainer, lockfile

20%

100

20.0 pts

Maintenancestaleness, pinning, CI

15%

100

15.0 pts

TransparencySKILL.md, perms, README

15%

100

15.0 pts

Communityinstalls, verify, response

15%

100

15.0 pts

Findings & checks · 2 flagged

Securityscore 76 · 2 findings

MEDIUMInstruction telling the agent not to ask for approvalSS-SKILL-INJECT-DONT-ASK-01 · Prompt injection · docs/competitors.md

MEDIUM — it fires on intent; the real damage depends on the host agent's own approval-gating.

Why it matters

The text {match} tells the agent to skip the normal "ask the user first" gate. Used adversarially it removes the human-in-the-loop check before destructive or sensitive actions, turning a normally-gated agent into a fire-and-forget executor.

The exact value spotted

excerptdocs/competitors.md· markdown

420### 4. Security is the emerging blocker

421

422**What:** Research proves trivial prompt injection via skills. "Don't ask again" approvals c

… (10 chars elided on L422)

423**Why now:** Skills gained broad access to filesystems and APIs. Supply chain risk is real.

424**Timeline:** Now, accelerating as skills from untrusted sources proliferate.

Occurrences

1 occurrence · at L422

How to fix

Remove the approval-skipping instruction, or scope it narrowly to a specific safe, reversible action.

Delete blanket "don't ask / no need to confirm" directives from the skill.
If the skill is a genuine autonomous job, restrict the opt-out to a named non-destructive action rather than all actions.

Framework references

OWASPLLM01 ATLASAML.T0051

Trace & refs

ruleSS-SKILL-INJECT-DONT-ASK-01sha25615dae9807bd892c6…rubric 365aacaView on GitHub

MEDIUMInstruction telling the agent not to ask for approvalSS-SKILL-INJECT-DONT-ASK-01 · Prompt injection · docs/problems.md

MEDIUM — it fires on intent; the real damage depends on the host agent's own approval-gating.

Why it matters

The exact value spotted

excerptdocs/problems.md· markdown

110- Long skill files make comprehensive human review difficult — malicious instructions hide i

… (16 chars elided on L110)

111- Skills operate in coding environments with access to sensitive files (API keys, `.env`)

112- A benign "Don't ask again" approval carries over to related harmful actions, bypassing gua

… (7 chars elided on L112)

113- **26.1% of 42,447 skills from marketplaces contained vulnerabilities** (prompt injection,

… (60 chars elided on L113)

114

Occurrences

1 occurrence · at L112

How to fix

Remove the approval-skipping instruction, or scope it narrowly to a specific safe, reversible action.

Delete blanket "don't ask / no need to confirm" directives from the skill.
If the skill is a genuine autonomous job, restrict the opt-out to a named non-destructive action rather than all actions.

Framework references

OWASPLLM01 ATLASAML.T0051

Trace & refs

ruleSS-SKILL-INJECT-DONT-ASK-01sha25615dae9807bd892c6…rubric 365aacaView on GitHub

Supply chainscore 100 · 0 findings

✓—All supply chain checks passedNo findings in this category for the latest scan.pass

Maintenancescore 100 · 0 findings

✓—All maintenance checks passedNo findings in this category for the latest scan.pass

Transparencyscore 100 · 0 findings

✓—All transparency checks passedNo findings in this category for the latest scan.pass

Communityscore 100 · 0 findings

✓—All community checks passedNo findings in this category for the latest scan.pass

The primary manifest — the file an agent reads to learn what this artifact does.

README.md · 1 file

README.md2.6 KB · Markdown

Skill Mix

A management layer for AI agent skills — discover, install, scope, rate, and update skills across Cursor, Codex, and Claude Code.

All three tools follow the open Agent Skills standard (SKILL.md), but none solves the management problem: how to find the right skill, avoid duplicates, track where it came from, or know if it's any good.

Setup

Quick install Skill Mix, open the app, and launch the skill-picker modal:

npx -y skill-mix obra/superpowers

Classic installer:

curl -fsSL https://raw.githubusercontent.com/razbakov/skills-manager/main/scripts/install.sh | bash

Usage

Run skills to launch application

Open/focus the app and trigger the skill-picker from CLI:

Skill Mix desktop UI

Build DMG (macOS)

bun install
bun run dist:dmg

This outputs a DMG installer in release/.

Problem

See [Research](docs/research.md) for how tools handle skills today, [User Problems](docs/problems.md) for 14 documented pain points, [Landscape](docs/landscape.md) for existing directories and tools, and [Competitors](docs/competitors.md) for detailed analysis of current solutions and their gaps.

Naming collisions — Multiple skills share similar names or overlap in responsibility. Cursor has no detection; Codex shows both without merging; Claude Code overrides by scope precedence.
No development workflow — No way to test, enable, or disable skills without editing files by hand (Codex has config.toml, but no UI).
Unknown provenance — None of the three tools track where a skill came from, who wrote it, or how to update it.
Scattered collections — Skills live in 3–6 different directories per tool with no unified view.
No quality signals — No ratings, benchmarks, or reviews exist anywhere in the ecosystem.

Key Features

[Skill Registry](docs/registry.md): Search, browse, and install skills from a shared catalog with deduplication and conflict detection.
[Scope Management](docs/scopes.md): Enable or disable skills at three levels — personal, organization, and project — with clear precedence rules across all supported tools.
[Provenance Tracking](docs/provenance.md): Every skill records its source repo, author, version, and update channel.
[Quality Scores](docs/quality.md): Community and automated ratings for security, efficiency, and best practices, with benchmarks across models and harnesses.
[Development Mode](docs/dev-mode.md): GUI for authoring, testing, and previewing skills before publishing.

skill-mix— agent skill

Is skill-mix safe to install?

Skill Mix

Setup

Usage

Build DMG (macOS)

Problem

Key Features

Audit the pieces. Scan the whole. Decide.