Agent Economy ReportAgent Economy ReportSAFE version 1.0.4 · audited 2026-09-09
This is a documentation-only skill: every file is Markdown text that tells the agent how to structure a coding workflow (plan, execute, verify, track state). It contains no scripts, no network calls, and no code that runs on its own. The only shell commands shown are two harmless setup lines that create a folder and an empty file at ~/code/memory.md, used to store coding preferences the user explicitly asks to save. The guidance repeatedly tells the agent to pause and ask the user before each step and not to touch files outside ~/code/ and the user's project. Two web addresses appear (clawhub.ai and clawic.com) but only as human-readable reference links in the skill card and homepage field, not as endpoints the agent contacts. Nothing in the package tries to read credentials, spend money, install software, or persist itself; the one modest caveat is that a user could ask the agent to save sensitive details into the local memory file, which the skill card itself flags.
| Declared purpose | "Coding workflow with planning, implementation, verification, and testing for clean software development" (SKILL.md line 6), including optional storage of user-stated coding preferences in ~/code/memory.md. |
| Observed behavior | Pure prompt/reference material. Instructs the agent to read ~/code/memory.md if present, break work into verifiable steps, wait for user approval between steps, verify with tests/screenshots, and append explicitly requested preferences to ~/code/memory.md. Suggests one setup command (mkdir -p ~/code; touch ~/code/memory.md). No network requests, no code execution, no credential access, no package installs, no persistence hooks. |
| Verdict, rules only | SAFE |
| Verdict, AI | SAFE |
| Final verdict | SAFE (never better than either pass) |
| capability | vs purpose | detail |
|---|---|---|
| filesystem_read | needed | Instructs reading ~/code/memory.md and the bundled reference files (SKILL.md line 54: "Read `~/code/memory.md` for user's stated preferences if it exists."). |
| filesystem_write | needed | Creates ~/code and ~/code/memory.md and appends user-requested preferences (memory-template.md lines 7-8; SKILL.md line 76). Scope is limited to the home ~/code directory and the user's project. |
| shell | needed | Only two benign shell lines are suggested in a bash code block: `mkdir -p ~/code` and `touch ~/code/memory.md` (memory-template.md lines 6-8). No other commands, no piping to shell. |
| severity | finding |
|---|---|
| info | Creates a directory in the user's home folder The only shell activity suggested by the skill is creating ~/code and an empty memory.md file. Both are benign and confined to a dedicated folder. memory-template.md:7 mkdir -p ~/code |
| low | Persists user-supplied text to a local file The skill writes preferences into ~/code/memory.md when the user asks. If a user asks it to remember sensitive project details or secrets, those would sit in plaintext on disk. The package's own skill-card.md (lines 19-20) discloses this risk and suggests reviewing the file. SKILL.md:76 | "Remember I prefer X" | Add to memory.md | |
| info | External URLs appear only as documentation links clawic.com (SKILL.md line 5) and clawhub.ai (skill-card.md line 7) are listed as homepage/publisher references. No code or instruction in the package fetches these or any other address; SKILL.md line 100 states the skill makes no network requests, consistent with the file contents. SKILL.md:5 homepage: https://clawic.com/skills/code |
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 |
| clawic.com | not in the service index | SKILL.md:5 |
| 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 |
|---|---|---|
SKILL.md | 120 | ad146d4d2a3ae667… |
_meta.json | 6 | 85eee13d53d9b8e8… |
criteria.md | 48 | 20d5332856656e57… |
execution.md | 42 | da81c0a72ef92408… |
memory-template.md | 38 | c46139090925509a… |
planning.md | 31 | 25675d904cc7d84f… |
skill-card.md | 38 | f2131fcc01a57ad5… |
state.md | 60 | 3e2451947af895ce… |
verification.md | 39 | 350148370baf9823… |
For agents
JSON: https://agenteconomy.report/k/code.audit.json · badge: https://agenteconomy.report/k/code.audited.svg ·
skill rating: /k/code · 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).