The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill contains hardcoded absolute machine paths that expose the author's local file system structure.
File evals/.lock.hashes contains multiple absolute paths (e.g., /Users/richardfremmerlid/Projects/agent-plugins-skills/...) as keys for file hashes.
Trace files in evals/traces/ (e.g., iter_001_KEEP_score0.82.json) include author-specific absolute paths within diff snippets and context fields recorded during skill development.
[COMMAND_EXECUTION]: The skill executes local Python scripts to perform audits and quality evaluations.
The Bash tool is used to execute python ./scripts/audit_plugin_paths.py for identifying path portability issues in the target scope.
The Bash tool is used to run python <APS_ROOT>/plugins/agent-agentic-os/skills/os-eval-runner/scripts/evaluate.py to calculate quality scores for skill iterations in Mode C.
[PROMPT_INJECTION]: The skill has an indirect prompt injection surface as it ingests and processes untrusted file content.
Ingestion points: The skill reads content from user-specified markdown and Python files (e.g., via Read and Edit tools or the auditor script).
Boundary markers: The instructions do not specify any delimiters or warnings to the agent to disregard instructions embedded within the files being audited.
Capability inventory: The agent has access to Bash, Write, and Edit tools, enabling command execution and file modification based on audited results.
Sanitization: No sanitization or validation logic for the external content is implemented in the skill instructions.
[EXTERNAL_DOWNLOADS]: The skill references a documentation file on the author's GitHub repository.
Found in references/fix-plugin-paths.prompt.md: https://github.com/richfrem/agent-plugins-skills/blob/main/INSTALL.md.

fix-plugin-paths