Agent Economy ReportAgent Economy ReportSAFE version 1.0.1 · audited 2026-09-09
This is a documentation-only skill: it contains no code, no scripts, and no executable commands. It is a reference sheet telling the agent how to format JSON payloads for an existing 'discord' tool that Clawdbot already provides (send/edit/delete messages, react, upload emojis and stickers, run polls, manage threads and pins, search, read member/role/channel info, and moderation actions like timeout/kick/ban). It also includes a writing-style guide for Discord messages. No network destinations are contacted, no credentials are read or transmitted, and no files are written by the skill itself; the only URL present is the publisher's own listing page in the skill card. The real risk is not in this package but in the underlying Discord tool's power: reading and deleting messages, uploading local files (file:///...) to Discord, changing roles, and banning users. The documentation notes that role and moderation groups are disabled by default and can be gated via discord.actions.*.
| Declared purpose | Guidance for controlling Discord from Clawdbot via the built-in discord tool: messaging, reactions, stickers/emojis, polls, threads, pins, search, server info, and moderation. |
| Observed behavior | Pure Markdown instruction content plus metadata. It documents action names and JSON parameter shapes and gives message-style advice. No shell execution, no HTTP requests, no environment variable access, no obfuscation, no persistence, and no attempt to override the agent's other rules. |
| Verdict, rules only | SAFE |
| Verdict, AI | SAFE |
| Final verdict | SAFE (never better than either pass) |
| capability | vs purpose | detail |
|---|---|---|
| network | needed | No network code exists in the package, but the documented actions cause the host's discord tool to call the Discord API with the configured bot token (SKILL.md line 10). |
| filesystem_read | needed | Documented media actions accept local paths such as 'file:///tmp/party.png' (SKILL.md lines 70, 87, 164), so the underlying tool can read local files and upload them to Discord. No file reading is performed by this package itself. |
| credentials | needed | Relies on the pre-existing Discord bot token held by Clawdbot (SKILL.md line 10). The skill does not read, print, or forward the token. |
| other | needed | Documents privileged Discord server operations: message deletion, role add/remove, and timeout/kick/ban. These are declared and disabled by default per lines 135-136. |
| severity | finding |
|---|---|
| info | Uses the host's existing Discord bot token The skill does not supply or handle credentials itself; it depends on a bot token already configured in the agent. The token's Discord permissions determine the real blast radius. SKILL.md:10 The tool uses the bot token configured for Clawdbot. |
| low | Destructive moderation actions documented, disabled by default The skill documents timeout, kick, and ban operations. It states these are off by default and gated through discord.actions.*, which is the appropriate default, but an operator who enables them gives the agent power to remove or silence server members. SKILL.md:136 - `moderation` (timeout/kick/ban, default `false`) |
| low | Message deletion and editing are in the default-enabled 'messages' group Unlike roles and moderation, the messages group is enabled by default (line 10 states groups default to enabled except roles/moderation), so the agent can edit or delete Discord messages without an extra opt-in. Irreversible for the affected content. SKILL.md:183 "action": "deleteMessage", |
| medium | Local files can be uploaded to Discord Documented media parameters accept local filesystem paths. If the agent is persuaded to pick a sensitive path, the underlying tool would post that file's contents into a Discord channel or DM. This is inherent to the declared upload feature, not hidden behavior, and the skill card lists it as a known risk. SKILL.md:164 "mediaUrl": "file:///tmp/audio.mp3" |
| low | Reads third-party chat content into the agent context readMessages and searchMessages pull arbitrary user-authored Discord text into the agent's context. Untrusted chat text can contain instructions; operators should not let Discord content drive privileged actions without confirmation. SKILL.md:141 "action": "readMessages", |
| info | Only external domain is the registry itself The single domain referenced anywhere in the package is clawhub.ai, used for publisher and skill listing links. No third-party or hidden endpoints appear. skill-card.md:7 [steipete](https://clawhub.ai/user/steipete) <br> |
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:7 |
| Binaries invoked | none |
| Environment variables read | none |
| Hard-coded wallets | none |
| Pipes a download to a shell | 0 |
| eval / exec / subprocess | 0 |
| base64 blobs | 0 |
| File writes | 1 |
| Persistence | 0 |
| file | lines | sha256 |
|---|---|---|
SKILL.md | 369 | 9ea23563b30f14bb… |
_meta.json | 6 | e70fa6c15e00337c… |
skill-card.md | 42 | ad76637997ac7570… |
For agents
JSON: https://agenteconomy.report/k/discord.audit.json · badge: https://agenteconomy.report/k/discord.audited.svg ·
skill rating: /k/discord · 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).