7skill

WarnAudited by ClawScan on May 18, 2026.

Overview

The PRD skill is mostly document-focused, but its references promote running an unattended coding agent with permissions bypassed.

Use this skill for drafting PRDs, but avoid the unattended `--dangerously-skip-permissions` workflow. If you want an agent to implement PRD stories, run it in a branch or worktree, require review between stories, and verify the package metadata mismatch before installing.

Findings (4)

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.

What this means

If followed, the agent could make broad project changes without normal permission prompts or human approval.

Why it was flagged

This recommends running a coding-agent CLI with permission checks skipped while allowing it to implement tasks, run checks, and update project files.

Skill content
claude --print --dangerously-skip-permissions "Read prd.json, find first story where passes=false, implement it, run checks, update passes=true if successful"
Recommendation

Do not use the permission-skipping mode for routine work. Run one story at a time, keep human review enabled, and use a sandbox or git worktree with explicit approvals for writes, commands, and commits.

What this means

A coding agent could continue operating repeatedly until manually stopped, compounding mistakes across the repository.

Why it was flagged

The documented workflow is an unbounded unattended loop; the shell loop itself has no iteration limit or review checkpoint.

Skill content
## Unattended Agentic Loop
while :; do ... done
Recommendation

Replace the infinite loop with a bounded workflow, explicit stop checks, and per-story human approval before continuing.

What this means

Incorrect or malicious content in progress.txt could steer later agent actions, and the file may accumulate implementation details.

Why it was flagged

The workflow uses a persistent local notes file as future agent context, which is purpose-aligned but can retain sensitive project details or be poisoned if edited by untrusted parties.

Skill content
Read `progress.txt` (check Codebase Patterns first) ... Append to `progress.txt` after each iteration (never replace)
Recommendation

Keep progress.txt project-local, review it before reuse, avoid storing secrets, and restrict who can edit it.

What this means

The package identity and provenance are harder to verify, even though no executable code is included.

Why it was flagged

These embedded identifiers do not match the supplied registry metadata, which lists a different owner ID, slug `excel-skill02`, and version `1.0.0`.

Skill content
"ownerId": "kn7f5p8fe35nfwgpfykcrswetd7ytn4n", "slug": "prd", "version": "2.0.5"
Recommendation

Verify the intended publisher and version before installing, and ask the publisher to align registry metadata with the packaged metadata.