ralph-loops@1.0.2
CAUTION· install onlyPublisher claims: "Runs autonomous iterative AI loops for requirements, planning, or building phases using structured prompts and fresh context per iteration." Observed in sandbox: reads credential files.
Coverage
Bundled scripts ran with no arguments; no agent task scenario for this bundle.- Tested
- 8 of 10 bundled scripts run with no arguments (the rest were not run: per-bundle cap), with decoy credentials in the environment · every file read, file written and host contacted by what ran
- Not tested
- use of the skill by an agent
Is [email protected] safe to install?
KillerSkills ran it in the KillerSkills honeypot sandbox (2 scenarios; bundled scripts ran with no arguments; no agent task scenario for this bundle) on 2026-11-19 and scored it 78/100 (Caution). It is not known to be malicious, but it has behaviour or code a security policy should decide on before install. Notable findings: Instructs the agent to modify its persistent memory or instruction files. Summary: reads credential files.
Behavior trace
Honeypot run · 16 s · 2 scenariosStatic findings
* @param {string} sessionKey - Session key to killFix: Build commands from fixed arguments; do not interpolate untrusted strings into a shell.* @returns {Object} Result of kill operationFix: Build commands from fixed arguments; do not interpolate untrusted strings into a shell.- **Add path restrictions:** In AGENTS.md, note which directories are in-scopeFix: Remove the text that addresses the agent rather than the reader; describe behavior, do not instruct.chmod +x loop.shFix: Request the narrowest permission that works; replace wildcards with the exact paths or hosts.* POST /api/loops/:id/kill - Kill a running loopFix: Build commands from fixed arguments; do not interpolate untrusted strings into a shell.execSync(`kill ${pid} 2>/dev/null`);Fix: Build commands from fixed arguments; do not interpolate untrusted strings into a shell.throw new Error(data.error || 'Failed to kill session');Fix: Build commands from fixed arguments; do not interpolate untrusted strings into a shell.console.log(` POST /api/loops/:id/kill - Kill session`);Fix: Remove the call, or declare the host in frontmatter and explain in the README what is sent and why.…rn. 9999999999. When you learn something new about how to run the application, update @AGENTS.md using a subagent but keep it brief. For example if you run commands multiple ti…Fix: Remove the text that addresses the agent rather than the reader; describe behavior, do not instruct.// Add click handlers for kill buttonsFix: Build commands from fixed arguments; do not interpolate untrusted strings into a shell.console.log(`\n⛔ Check command signaled stop: ${output.trim()}`);Fix: Remove the call, or declare the host in frontmatter and explain in the README what is sent and why.console.log(` Output: ${output.slice(0, 200)}`);Fix: Remove the call, or declare the host in frontmatter and explain in the README what is sent and why.console.log(` POST /api/loops/:id/kill - Kill session`);Fix: Build commands from fixed arguments; do not interpolate untrusted strings into a shell.console.log('\n📝 Raw output:', output.slice(0, 300));Fix: Remove the call, or declare the host in frontmatter and explain in the README what is sent and why.this.showError('Failed to kill session: ' + error.message);Fix: Build commands from fixed arguments; do not interpolate untrusted strings into a shell.Cost to the agent
Tokens estimated at four characters each. Skills load their instruction text when triggered; the figure is what that costs.
Known vulnerabilities
Vulnerability advisories (OSV) cover npm and PyPI packages; this registry has no advisory database.
What would make this Verified
2 points to 80 · weights: behavior 45%, static 25%, prompt 15%, publisher 15%- Reads credential files≈ +27 on the totalThe sandbox saw reads under ~/.ssh, ~/.aws, ~/.npmrc, ~/.config or similar. Read only files inside the project, and only the ones the task needs. Seen: ~/.npmrc.
- Unclaimed publisher≈ +8 on the totalClaim qlifebot-coder: a verified publisher gets +50 on publisher trust. It is free and usually instant with GitHub sign-in.
- Touches files outside the project≈ +7 on the totalKeep reads and writes inside the working directory, or declare the exact paths in `filesystem:` and explain why in the README.
- Process kill command≈ +2 on the totalBuild commands from fixed arguments; do not interpolate untrusted strings into a shell.
- File permission modification command≈ +2 on the totalRequest the narrowest permission that works; replace wildcards with the exact paths or hosts.
- File type combined with upload action≈ +2 on the totalRemove the call, or declare the host in frontmatter and explain in the README what is sent and why.