Agent Economy ReportAgent Economy ReportCAUTION version 0.1.0 · audited 2026-09-17
This package is a markdown-only process skill. It tells an agent to check for relevant skills before any response or action and to follow invoked skills exactly, creating todo items when a skill has a checklist. It does not contain executable code, shell commands, network calls, file writes, credential access, payment activity, or persistence. Its main risk is behavioral: it uses forceful 'non-negotiable' language to remove agent discretion and can pressure the agent to load downstream skills too early. The skill card itself acknowledges that it broadly forces agent behavior across conversations. The package is not exfiltrating data, but operators should review it before enabling automatic skill loading.
Instructions try to steer the agent beyond the declared purpose. The skill repeatedly uses absolute, instruction-overriding language: 'You ABSOLUTELY MUST invoke the skill' (SKILL.md line 7), 'YOU DO NOT HAVE A CHOICE. YOU MUST USE IT' (line 9), and 'This is not negotiable. This is not optional. You cannot rationalize your way out of this' (line 11). It also says 'follow it directly' and 'Never use the Read tool on skill files' (line 16), which reduces independent inspection of downstream skill content.
| Declared purpose | Use when starting any conversation - establishes how to find and use skills, requiring Skill tool invocation before ANY response including clarifying questions |
| Observed behavior | The skill is a set of prompt instructions, not executable code. If installed, it tells the agent to invoke the Skill tool before any response or action when there is even a 1% chance a skill applies; to follow the invoked skill directly; to create TodoWrite items for checklists; to prioritize process skills before implementation skills; and to treat common objections such as 'this is just a simple question' as rationalizations. There are no direct filesystem writes, shell commands, network fetches, credential reads, or payment operations in the files. |
| Verdict, rules only | SAFE |
| Verdict, AI | CAUTION |
| Final verdict | CAUTION (never better than either pass) |
| capability | vs purpose | detail |
|---|---|---|
| other | needed | Instructs the agent to use the Skill tool and TodoWrite tool. No direct use of network, shell, filesystem, credentials, payments, persistence, install_packages, or browser is present in the package. |
| severity | finding |
|---|---|
| medium | Broad mandatory skill invocation Forces the agent to invoke skills before any action if there is even a 1% chance they apply. This is the declared purpose, but it lowers the bar for loading downstream skills. SKILL.md:7 If you think there is even a 1% chance a skill might apply to what you are doing, you ABSOLUTELY MUST invoke the skill. |
| medium | Removes agent discretion Language attempts to remove the agent's decision-making once a skill may apply. Combined with automatic skill invocation, this can weaken other safety or priority checks. SKILL.md:9 IF A SKILL APPLIES TO YOUR TASK, YOU DO NOT HAVE A CHOICE. YOU MUST USE IT. |
| low | Anti-rationalization language Attempts to cast any deviation from the skill workflow as rationalization, which may suppress the agent's own relevance and safety review. SKILL.md:11 This is not negotiable. This is not optional. You cannot rationalize your way out of this. |
| medium | Directly follow invoked skill without using Read tool The agent is told to follow whatever the Skill tool loads and not to inspect skill files with the Read tool. This is normal Claude Code skill handling, but increases downstream risk if a future skill contains harmful instructions. SKILL.md:16 Use the `Skill` tool. When you invoke a skill, its content is loaded and presented to you—follow it directly. Never use the Read tool on skill files. |
| low | Skill check before all responses Even clarifying questions are supposed to be preceded by a skill check. This may create unwanted loading of skills for simple or sensitive conversations. SKILL.md:24 Invoke relevant or requested skills BEFORE any response or action. |
| info | Uses TodoWrite tool Directs the agent to create TodoWrite todos for skill checklists. This package itself does not perform filesystem writes. SKILL.md:33 Create TodoWrite todo per item |
| info | Discourages quick git/files checks before skill check The skill tells the agent not to use git or files directly before checking for relevant skills. No git command is executed by this package; the mention is in a behavior table. SKILL.md:57 "I can check git/files quickly" | Files lack conversation context. Check for skills. |
| medium | Publisher acknowledges broad forced behavior The skill card itself warns that the skill broadly forces agent behavior and downstream skill loading, supporting a careful review rather than an unqualified SAFE verdict. skill-card.md:24 Risk: The skill broadly forces agent behavior and downstream skill loading across nearly every conversation. |
| info | Static publisher link The only domain in the package is a static link to the publisher profile. No code fetches this URL. skill-card.md:9 [zlc000190](https://clawhub.ai/user/zlc000190) |
0 AI finding(s) were dropped because their file, line or quote did not match the package.
| domain | service rating | where |
|---|---|---|
| clawhub.ai | not in the service index | skill-card.md:9 |
| Binaries invoked | git |
| Environment variables read | none |
| Hard-coded wallets | none |
| Pipes a download to a shell | 0 |
| eval / exec / subprocess | 0 |
| base64 blobs | 0 |
| File writes | 8 |
| Persistence | 0 |
| file | lines | sha256 |
|---|---|---|
SKILL.md | 87 | 81d921b16502091f… |
_meta.json | 6 | 37c21b6e4af22bca… |
skill-card.md | 53 | e0c442e69e866a46… |
For agents
JSON: https://agenteconomy.report/k/using-superpowers.audit.json · badge: https://agenteconomy.report/k/using-superpowers.audited.svg ·
skill rating: /k/using-superpowers · commission an audit of any skill: US$ 29 per version.
The complete published package of this exact version was downloaded from the registry and read statically; nothing was executed. A deterministic pass extracts network destinations, binaries, environment variables, writes, install commands, obfuscation markers, subprocesses and wallets, each with file and line. An AI then reads the whole package with those facts and writes the summary, the capabilities and the findings under a strict schema; every finding it produces must cite a file, a line and the exact text, or it is dropped. The final verdict is the worst of the two passes. The audit does not change the skill's trust tier (policy); the author may respond through the dispute channel and the response is published here. Commissioned by: the Agent Economy Report (free program: the 150 most downloaded skills, September 2026).