registry / @ai-dev-methodologies/rlp-desk / 0.26.0
@ai-dev-methodologies/rlp-desk@0.26.0
Fresh-context iterative loops for Claude Code — autonomous task completion with independent verification
AI Security Review
scanned 8d ago · by lpm-firewall-aiThe automatic postinstall performs first-party Claude Code extension setup by copying the package’s own command and runtime into user-level Claude directories. Source inspection found no separate concrete attack, so this warrants a warning rather than a publish block.
Static reason
High-risk behavior combination matched malicious policy.; previous stored version diff introduced dangerous source
Trigger
Installing the package invokes its declared postinstall hook.
Impact
Installation adds a package-owned Claude Code command and runtime that can be used in later sessions.
Mechanism
The hook copies manifest-listed package assets and the Node runtime into ~/.claude/commands and ~/.claude/ralph-desk.
AI rationale
The automatic postinstall performs first-party Claude Code extension setup by copying the package’s own command and runtime into user-level Claude directories. Source inspection found no separate concrete attack, so this warrants a warning rather than a publish block.
Evidence
package.jsonscripts/postinstall.jsscripts/install-manifest.js
Decision evidence
public snapshotAI called this Suspicious at 90.0% confidence as Dangerous Capability with medium false-positive risk.
Evidence for warning
- The manifest declares a postinstall hook that runs the package installer.
- The install manifest maps this package’s own Claude Code command and runtime files into the user’s Claude directories.
- The hook copies the manifest assets and the package’s Node runtime into those locations during installation.
- The campaign initializer migrates legacy project data and creates campaign files under the resolved project desk root.
- The runtime points to the installed package runner; the legacy path in the campaign loop identifies a project campaign verdict.
- The registry code stores campaign state metadata in a local file under the package’s desk directory.
Evidence against
- Inspected source does not establish unrelated agent settings mutation, credential transfer, or network exfiltration.
Behavioral surface
ChildProcessCryptoDynamicRequireEnvironmentVarsFilesystem
HighEntropyStrings
Source & flagged code
8 flagged · loading sourcepackage.jsonView file
•scripts.postinstall = node scripts/postinstall.js
High
Install Time Lifecycle Scripts
Package defines install-time lifecycle scripts.
package.jsonView on unpkg•scripts.postinstall = node scripts/postinstall.js
Medium
Ambiguous Install Lifecycle Script
Install-time lifecycle script is not statically allowlisted and needs review.
package.jsonView on unpkgscripts/postinstall.jsView file
33function ensureDir(dirPath) {
L34: fs.mkdirSync(dirPath, { recursive: true });
L35: }
...
L40: // Security review v5.7 follow-up: lstatSync first; SKIP symlinks entirely so
L41: // a hostile symlink (e.g., ~/.claude/ralph-desk/foo -> /etc/passwd) cannot
L42: // be chmod'd via unlockTree's chmodSync (which follows symlinks).
...
L95: // ONLY if a developer ran sync from an installed copy back to the source —
L96: // which is forbidden by CLAUDE.md. But re-running install over an existing
L97: // installed file (e.g., npm i again) does NOT need re-injection because
L98: // copyFileSync replaced the file with the source contents. The check below
L99: // is defensive — only inject when the file does not already start with a
...
L113: const tailBuf = original.subarray(newlineIdx + 1);
Findings
1 Critical4 High5 Medium4 Low
CriticalAi Agent Control Hijackscripts/postinstall.js
HighInstall Time Lifecycle Scriptspackage.json
HighTrigger Reachable External Ai Agent Control Surface Mutationscripts/postinstall.js
HighSemantic Analysis Limitedsrc/node/run.mjs#virtual:normalized:round1
HighPrevious Version Dangerous Deltasrc/node/reporting/campaign-reporting.mjs
MediumAmbiguous Install Lifecycle Scriptpackage.json
MediumDynamic Requirescripts/postinstall.js
MediumEnvironment Vars
MediumShips Build Helperinstall.sh
MediumStructural Risk Force Deep Review
LowNon Install Lifecycle Scripts
LowScripts Present
Affected versions and remediation
This report applies to @ai-dev-methodologies/rlp-desk@0.26.0.
See all versions of @ai-dev-methodologies/rlp-desk for other recorded verdicts.
- Review the evidence and your use of @ai-dev-methodologies/rlp-desk@0.26.0 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.