@runcanary/cli@0.1.9
Canary CLI — authenticate and onboard your repo to Canary from the terminal.
AI Security Review
scanned 1mo ago · by lpm-firewall-aiThe observed behavior is an explicit CLI capability rather than an install-time attack, but it modifies AI-agent instruction surfaces by default during login. It warrants a warning under the agent configuration policy.
Static reason
One or more suspicious static signals were detected.
Trigger
A user runs the canary login command without disabling agent skills.
Impact
This adds package-provided instructions to supported coding-agent environments.
Mechanism
The login flow detects existing agent directories and writes or refreshes a Canary SKILL.md file below their skills directories.
AI rationale
The observed behavior is an explicit CLI capability rather than an install-time attack, but it modifies AI-agent instruction surfaces by default during login. It warrants a warning under the agent configuration policy.
Evidence
dist/canary.js~/.claude/skills/canary/SKILL.md~/.agents/skills/canary/SKILL.md~/.codex/skills/canary/SKILL.md
Decision evidence
public snapshotAI called this Suspicious at 91.0% confidence as Dangerous Capability with low false-positive risk.
Evidence for warning
- The package exposes a command-line entry point rather than an automatic install hook.
- The login flow can create or refresh Canary skill files beneath detected Claude, Agents, and Codex directories.
- This agent configuration behavior is user-invoked through login and can be disabled with a command option.
Evidence against
- No preinstall, install, or postinstall hook is declared.
- The skill writer checks for symbolic links before writing.
Behavioral surface
ChildProcessEnvironmentVarsNetwork
HighEntropyStringsMinifiedUrlStrings
Source & flagged code
5 flaggeddist/canary.jsView file
12(Did you mean one of ${i.join(", ")}?)`:i.length===1?`
L13: (Did you mean ${i[0]}?)`:""}ln.suggestSimilar=ar});var gn=ge(hn=>{var cr=Fe("node:events").EventEmitter,It=Fe("node:child_process"),Z=Fe("node:path"),Et=Fe("node:fs"),T=Fe("node:pr...
L14: - specify the name in Command constructor or using .name()`);return t=t||{},t.isDefault&&(this._defaultCommandName=e._name),(t.noHelp||t.hidden)&&(e._hidden=!0),this._registerComma...
High
12(Did you mean one of ${i.join(", ")}?)`:i.length===1?`
L13: (Did you mean ${i[0]}?)`:""}ln.suggestSimilar=ar});var gn=ge(hn=>{var cr=Fe("node:events").EventEmitter,It=Fe("node:child_process"),Z=Fe("node:path"),Et=Fe("node:fs"),T=Fe("node:pr...
L14: - specify the name in Command constructor or using .name()`);return t=t||{},t.isDefault&&(this._defaultCommandName=e._name),(t.noHelp||t.hidden)&&(e._hidden=!0),this._registerComma...
...
L24: Expecting one of '${i.join("', '")}'`);let r=`${e}Help`;return this.on(r,s=>{let o;typeof t=="function"?o=t({error:s.error,command:s.command}):o=t,o&&s.write(`${o}
L25: `)}),this}_outputHelpIfRequested(e){let t=this._getHelpOption();t&&e.find(r=>t.is(r))&&(this.outputHelp(),this._exit(0,"commander.helpDisplayed","(outputHelp)"))}};function pn(n){r...
High
Same File Env Network Execution
A single source file combines environment access, network access, and code or shell execution; review context before blocking.
dist/canary.jsView on unpkg · L12Findings
4 High4 Medium4 Low
HighChild Processdist/canary.js
HighSame File Env Network Executiondist/canary.js
HighCommand Output Exfiltrationdist/canary.js
HighTrigger Reachable Command Output Exfiltrationdist/canary.js
MediumNetwork
MediumEnvironment Vars
MediumStructural Risk Force Deep Review
MediumSuspicious Lifecycle Evidencepackage.json
LowNon Install Lifecycle Scripts
LowScripts Present
LowHigh Entropy Strings
LowUrl Strings
Affected versions and remediation
This report applies to @runcanary/cli@0.1.9.
See all versions of @runcanary/cli for other recorded verdicts.
- Review the evidence and your use of @runcanary/cli@0.1.9 before allowing it. Restrict the permissions described in this report.
- Choose an independently verified alternative or release. This report does not establish that other versions are safe.
Evidence last updated: .
Think this verdict is wrong? Dispute it on GitHub or read how reports are produced.