The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute a Python script at .claude/scripts/analyze_errors.py and various shell scripts located in .claude/hooks/. Because the source code for these scripts is not provided within the skill package, their behavior cannot be verified, posing a risk of arbitrary command execution.
[EXTERNAL_DOWNLOADS]: The documentation encourages the global installation of the portless utility using npm i -g portless. This introduces an external dependency from a third-party registry that is not audited as part of the skill.
[DATA_EXFILTRATION]: The skill implements an error collector that logs tool names, input commands, and error messages to .claude/logs/errors.jsonl. This mechanism may capture and store sensitive information, such as credentials or private data included in failed command arguments, in a file that the agent is subsequently instructed to read.
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection by processing untrusted data from error logs.
Ingestion points: The agent reads and analyzes historical error data from .claude/logs/errors.jsonl.
Boundary markers: The skill lacks delimiters or protective instructions to prevent the agent from following malicious commands embedded within captured error messages.
Capability inventory: The skill utilizes Read, Bash, and Grep tools, which could be exploited if an attacker-controlled process generates error messages designed to manipulate the agent's logic.
Sanitization: There is no evidence of filtering or sanitizing error messages or tool inputs before they are processed by the analysis scripts.

errors