3skill
Security checks across static analysis, malware telemetry, and agentic risk
Overview
This PRD helper is mostly documentation, but it includes an unattended coding-agent loop that bypasses permissions and can keep editing or committing code.
Use this skill as a PRD template only unless you intentionally want autonomous code execution. Avoid the documented --dangerously-skip-permissions loop, set strict iteration limits, review every diff and commit manually, and keep progress.txt free of untrusted instructions.
Static analysis
No static analysis findings were reported for this release.
VirusTotal
VirusTotal findings are pending for this skill version.
Risk analysis
Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.
If followed, an agent could modify the codebase and run commands without normal permission checks.
This recommends running a coding agent with permission prompts disabled while it implements work and updates project files, which is high-impact tool use without a clear approval boundary.
claude --print --dangerously-skip-permissions "Read prd.json, find first story where passes=false, implement it, run checks, update passes=true if successful"
Do not use the dangerous permission-bypass mode for routine work; require explicit approval for file changes, commands, and commits, and review diffs before marking stories complete.
The agent may continue operating, consuming resources, or attempting changes beyond the user’s expected stopping point.
The documented unattended loop is unbounded at the shell level and can keep launching autonomous agent runs after the intended task unless externally stopped.
while :; do
claude --print --dangerously-skip-permissions \
"Read prd.json, find first story where passes=false, implement it, run checks, update passes=true if successful"
doneReplace infinite loops with a bounded runner, explicit stop checks, maximum iteration/time limits, and human review between stories.
Incorrect or malicious content placed in progress.txt could influence later agent changes.
The workflow creates a persistent project note that future agent iterations are told to read and trust for codebase patterns.
Read `progress.txt` (check Codebase Patterns first) ... Append progress to `progress.txt`
Keep progress.txt under review, treat it as untrusted project context, and avoid letting it override explicit user instructions or repository policy.
This does not show malicious behavior, but it makes the package identity and version lineage less clear.
The internal metadata differs from the supplied registry metadata, which lists a different owner ID, slug, and version.
"ownerId": "kn7f5p8fe35nfwgpfykcrswetd7ytn4n", "slug": "prd", "version": "2.0.5"
Confirm the intended publisher and version before relying on the skill, especially if it is installed from an unknown source.