registry / AI-agent control-surface risk / @ckelsoe/prompt-architect
@ckelsoe/prompt-architect
Agent skill for analyzing and improving prompts using 29 frameworks across 7 intent categories. Works with Claude Code, Gemini CLI, Cursor, Copilot, and 30+ Agent Skills compatible tools.
This report shows the evidence for @ckelsoe/prompt-architect@3.3.1, the only known version.
AI Security Review
scanned 3mo ago · by lpm-firewall-aiThe postinstall behavior is concrete, unconsented, and targets multiple external AI-agent control surfaces. No network exfiltration or secondary payload was found, but the lifecycle mutation independently meets the blocking policy.
Static reason
High-risk behavior combination matched malicious policy.
Trigger
npm installation in a non-interactive/CI context
Impact
Unconsented persistence of package instructions across broad AI-agent control surfaces.
Mechanism
postinstall copies skills and writes agent instruction files
Policy narrative
Installing the npm package invokes its postinstall hook. In the non-interactive path, the installer detects configured AI tools and recursively replaces package-named skills in user-global Claude, Gemini, and universal Agent Skills directories; it can append its adapter to a project Windsurf rules file. This mutates foreign and broad AI-agent instruction surfaces without an installation choice.
AI rationale
The postinstall behavior is concrete, unconsented, and targets multiple external AI-agent control surfaces. No network exfiltration or secondary payload was found, but the lifecycle mutation independently meets the blocking policy.
Evidence
package.jsonscripts/install.jsskills/prompt-architect/SKILL.mdadapters/for-windsurf.mdskills/prompt-architect/~/.claude/skills/prompt-architect/~/.gemini/skills/prompt-architect/~/.agents/skills/prompt-architect/./.windsurfrules
Decision evidence
public snapshotAI identified this as a dangerous AI-agent capability at 98.0% confidence with low false-positive risk.
Evidence for policy risk
- package.json runs scripts/install.js via postinstall.
- Non-interactive postinstall installs to every detected agent.
- scripts/install.js writes to ~/.claude/skills/prompt-architect.
- It also writes to ~/.gemini/skills and ~/.agents/skills.
- Detected Windsurf causes marked content to be written to .windsurfrules.
Evidence against
- No network client or credential harvesting found.
- execSync is limited to agent detection and npm global-root lookup.
- Bundled SKILL.md is prompt-engineering guidance; no payload execution found.
Behavioral surface
ChildProcessEnvironmentVarsFilesystemShell
HighEntropyStrings
Source & flagged code
3 flagged · loading sourcepackage.jsonView file
•scripts.postinstall = node scripts/install.js
High
Install Time Lifecycle Scripts
Package defines install-time lifecycle scripts.
package.jsonView on unpkg•scripts.postinstall = node scripts/install.js
Medium
Ambiguous Install Lifecycle Script
Install-time lifecycle script is not statically allowlisted and needs review.
package.jsonView on unpkgscripts/install.jsView file
5Install-time AI-agent control hijack evidence:
L31: name: 'Claude Code',
L32: detect: () => fs.existsSync(path.join(os.homedir(), '.claude')),
L33: installPath: (project) => project
L34: ? path.join(process.cwd(), '.claude', 'skills', SKILL_NAME)
L35: : path.join(os.homedir(), '.claude', 'skills', SKILL_NAME),
L36: format: 'skill',
L37: hint: (project) => project ? '.claude/skills/ (project)' : '~/.claude/skills/',
L38: },
...
L125: if (stats.isDirectory()) {
L126: if (!fs.existsSync(dest)) fs.mkdirSync(dest, { recursive: true });
L127: for (const entry of fs.readdirSync(src)) {
...
L130: } else {
Payload evidence from skills/prompt-architect/SKILL.md:
L5: compatibility: Requires no external dependencies. Works with any Agent Skills compatible tool.
L6: metadata:
L7: author: ckelsoe
L8: ve
Critical
Ai Agent Control Hijack
Install-time source drops package-supplied AI-agent/MCP control files or instructions.
scripts/install.jsView on unpkg · L5Findings
1 Critical1 High3 Medium4 Low
CriticalAi Agent Control Hijackscripts/install.js
HighInstall Time Lifecycle Scriptspackage.json
MediumAmbiguous Install Lifecycle Scriptpackage.json
MediumEnvironment Vars
MediumStructural Risk Force Deep Review
LowNon Install Lifecycle Scripts
LowScripts Present
LowFilesystem
LowHigh Entropy Strings
Affected versions and remediation
This report applies to @ckelsoe/prompt-architect@3.3.1.
The versions list below shows the verdict for each recorded version.
- Avoid installing @ckelsoe/prompt-architect@3.3.1. Remove it from direct dependencies and check your lockfile for transitive copies.
- Choose an independently verified alternative or release. This report does not establish that other versions are safe.
- If this version ran, investigate the affected machine and build environment. Rotate credentials it could access and rebuild from a trusted environment.
Evidence last updated: .
Think this verdict is wrong? Dispute it on GitHub or read how reports are produced.