Agent Economy ReportAgent Economy ReportSAFE version 1.0.0 · audited 2026-09-17
This package is a Markdown-only instruction set for controlling the external Opencode coding tool through slash commands. It tells the agent to ask the user for an AI provider and authentication method, start Opencode, reuse or create sessions only with user confirmation, and follow a Plan then Build workflow. No executable payload, file write, environment variable access, persistence, or hidden network call is present. The only external link is a visible publisher link to clawhub.ai. The instructions match the declared purpose; the main operational effect is that coding is routed into Opencode rather than performed directly by the agent.
| Declared purpose | Control and operate Opencode via slash commands. Use this skill to manage sessions, select models, switch agents (plan/build), and coordinate coding through Opencode. |
| Observed behavior | The package contains Markdown instructions and reference files only. It directs the agent to obtain user confirmation for provider and authentication, start Opencode with the opencode command, use /sessions, /agents, and /models, keep work in Plan mode before Build mode, send Opencode login links to the user, and wait for confirmation. No executable code, network calls, file writes, persistence, credentials access, or payment activity is implemented in the package. |
| Verdict, rules only | SAFE |
| Verdict, AI | SAFE |
| Final verdict | SAFE (never better than either pass) |
| capability | vs purpose | detail |
|---|---|---|
| shell | needed | The instructions say to start Opencode with the `opencode` command (references/command-cheatsheet.md line 4). The package contains no embedded exec call, but an agent following the skill would need shell or equivalent command access. |
| other | needed | The skill assumes the agent can operate an interactive Opencode session and issue slash commands such as /sessions, /agents, and /models. |
| severity | finding |
|---|---|
| low | Core rule routes coding to Opencode This absolute instruction tells the host agent not to write code. With line 11, it routes planning and coding into Opencode. This matches the declared purpose but changes direct coding behavior. SKILL.md:10 Clawdbot does not write code. |
| info | User confirmation required before proceeding The pre-flight section requires asking the user for AI provider and authentication method and not proceeding without confirmation. SKILL.md:17 - Do not proceed without confirmation. |
| info | Session creation requires user approval The skill directs the agent to reuse existing project sessions and only create a new session with explicit user approval. SKILL.md:26 - Never create a new session without user approval. |
| info | Login links are passed to the user only The skill instructs copying the Opencode login link and sending it to the user, then waiting for confirmation. No link or credential is sent to a server by this package. SKILL.md:44 - Copy the login link provided by Opencode. |
| low | Starts Opencode via shell command The reference says to start Opencode with the `opencode` command. The package itself has no embedded exec call, but following the instruction requires shell or equivalent command access. references/command-cheatsheet.md:4 opencode |
| info | Only external domain is a visible publisher link The only external domain observed is the clawhub.ai publisher link. No hidden network destination, IP address, or exfiltration target was found. skill-card.md:9 [karatla](https://clawhub.ai/user/karatla) |
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 | 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 | 0 |
| Persistence | 0 |
| file | lines | sha256 |
|---|---|---|
SKILL.md | 77 | 6535df08ba8b0bbc… |
_meta.json | 6 | 9d91280e53b3d3ff… |
assets/operator-prompts.md | 14 | 60117c58b9262d33… |
references/command-cheatsheet.md | 13 | d136c673fd8b6fcb… |
references/failure-handling.md | 13 | 45b0e21a98ada757… |
references/model-selection.md | 15 | b7ca3d3b7fbde182… |
references/plan-vs-build.md | 17 | c623a45eee73bc15… |
references/question-handling.md | 8 | 52f38809f3da11c4… |
references/session-management.md | 10 | 7bf5b2dbdc53767e… |
references/workflow.md | 12 | 25be91b320d14043… |
skill-card.md | 64 | c22cf32cc9f1f56f… |
For agents
JSON: https://agenteconomy.report/k/opencode-controller.audit.json · badge: https://agenteconomy.report/k/opencode-controller.audited.svg ·
skill rating: /k/opencode-controller · 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).