Agent Economy ReportAgent Economy ReportCAUTION version 1.0.0 · audited 2026-09-09
This package contains no executable code at all two markdown files (SKILL.md, README.md) and a metadata file. Its content is descriptive prose claiming the skill makes an AI agent more "proactive": remembering context, suggesting ideas unprompted, self-healing, and staying on mission. There are no scripts, no network addresses, no environment variables read, no files written, and no commands run; the only shell reference is the documented install command `clawhub install proactive-agent-lite`. Because nothing is implemented, installing it does nothing except add prompt text to the agent's context. The one thing a careful reader should notice is that the text encourages the agent to act and create value "without being asked," which is a nudge toward unprompted autonomous action for an agent that may hold credentials or spend money. It also lists none of the concrete mechanisms it advertises, so the described features (memory flush, self-healing, tuning parameters) do not exist in the package.
| Declared purpose | To turn AI agents from passive task-followers into proactive partners via memory architecture, reverse prompting, security hardening, self-healing and alignment patterns (lightweight version). |
| Observed behavior | Documentation-only package. No code, no shell execution, no network calls, no credential or filesystem access. Its only effect is to inject descriptive/behavioral prose into the agent's context; the advertised features have no implementation in the files provided. |
| Verdict, rules only | SAFE |
| Verdict, AI | CAUTION |
| Final verdict | CAUTION (never better than either pass) |
| capability | vs purpose | detail |
|---|---|---|
| other | needed | Prompt-text-only behavior modification. The package ships two markdown files and a metadata file; it declares no requirements and no install steps (SKILL.md line 8-9: "requires": {}, "install": []). |
| severity | finding |
|---|---|
| low | Encourages unprompted autonomous action The skill's whole payload is prompt text that pushes the agent to act on its own initiative. In an agent that holds credentials or can spend money, instructions to take action "without being asked" increase the chance of unrequested side effects. There is no code enforcing limits, so the actual risk depends entirely on the host agent's own guardrails. README.md:9 - Create value without being asked |
| low | Claims automatic behavior change on install with no implementation The package contains no scripts or configuration that could produce this behavior; the effect, if any, comes only from the agent reading this prose. Users should not expect the advertised safeguards ("Security Hardening", "safe defaults") to be technically enforced. SKILL.md:36 Simply install and the agent will automatically begin exhibiting proactive characteristics. |
| info | Advertised configuration does not exist in the package No configuration file, schema, or parameter is included anywhere in the package, so the documented tuning options are not present. This is a documentation/accuracy gap, not a security flaw. SKILL.md:49 The proactive behavior can be tuned through configuration parameters to match your preferred level of initiative and communication style. |
| info | Only shell reference is the documented install command The single bash block (README.md lines 16-18) contains `clawhub install proactive-agent-lite`, a normal registry install instruction. There is no download-and-execute, no piping to a shell, and no other command execution in the package. README.md:16 ```bash |
0 AI finding(s) were dropped because their file, line or quote did not match the package.
| domain | service rating | where |
|---|---|---|
| No network destination found in the code. | ||
| Binaries invoked | bash |
| Environment variables read | none |
| Hard-coded wallets | none |
| Pipes a download to a shell | 0 |
| eval / exec / subprocess | 0 |
| base64 blobs | 0 |
| File writes | 0 |
| Persistence | 0 |
| file | lines | sha256 |
|---|---|---|
README.md | 66 | b6af11904f4ae42a… |
SKILL.md | 49 | d8511a1ecd63fda4… |
_meta.json | 6 | 4ec09c98fafe08fb… |
For agents
JSON: https://agenteconomy.report/k/proactive-agent-lite.audit.json · badge: https://agenteconomy.report/k/proactive-agent-lite.audited.svg ·
skill rating: /k/proactive-agent-lite · 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).