pddl-skills
PassAudited by ClawScan on May 10, 2026.
Overview
This is a straightforward instruction-only PDDL planning helper with no evidence of hidden credentials, network access, persistence, or destructive behavior.
This skill appears safe for normal PDDL planning use. Before installing or using it, make sure any required Python planning libraries are installed from trusted sources, and only provide file paths you intend the agent to read or write.
Findings (2)
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 given the wrong path, the agent could read an unintended local file or overwrite an unintended output file.
The skill involves local file reads and writes via user-provided paths. This is expected for loading PDDL inputs and saving plan outputs, but users should keep paths scoped to intended benchmark and output files.
`load-problem(domain_path, problem_path)` ... `save-plan(plan_object, output_path)` ... `Writes a plan object to disk in standard PDDL plan format.`
Use explicit, project-local paths for domain, problem, and plan files, and review any save location before writing.
Functionality may depend on whatever local versions of these libraries are installed, which could affect reproducibility or security posture.
The skill depends on external planning libraries, while the provided install specification says this is instruction-only and does not pin or install those dependencies.
Uses `unified_planning.io.PDDLReader` ... Uses `unified_planning.shortcuts.OneshotPlanner`. Default planner: `pyperplan`.
Install dependencies from trusted package sources and pin versions in your own environment if reproducibility matters.