Security audit
Security checks for vulnerabilities and agentic risk
This is a local encoding and decoding helper whose command examples are mostly purpose-aligned, with some documentation quality issues users should notice before running file-changing commands.
Install only if you want a Markdown helper for common encoding, decoding, JWT inspection, and hashing tasks. Before running its shell examples, confirm the target input and avoid in-place file edits such as sed -i unless you have a backup; prefer local standard-library tools over any optional online hash service.
The activation guidance is broad and ambiguous, including generic uses like code generation, programming assistance, debugging, testing, and deployment. In an agent environment with exec/read/glob/grep tools, this can cause the skill to activate outside narrowly intended encoding tasks and lead the agent to run shell commands in contexts where the user did not specifically request this skill.
The coverage-keywords section massively broadens scope with fragmented and loosely constrained keywords such as code generation, debugging, deployment, and general developer workflows. This increases the chance of unintended activation and tool use, especially because the skill advertises exec capability and presents operational shell examples.
The examples include commands and workflows that can write binary output to files or otherwise alter local state, but the skill does not warn users or the agent about filesystem impact before using exec-driven shell commands. In an agent setting, omission of such warnings can normalize unsafe execution patterns and lead to accidental overwrites or unintended file creation during routine use.
The BOM removal example uses sed -i to modify files in place without any confirmation, backup, or warning. In an agent context with exec access, this is dangerous because it can irreversibly alter user files, and if applied to the wrong target or pattern, may cause data loss or corrupt important text assets.
L202 states the skill is based on local command-line tools and standard libraries and requires no additional API key. Immediately after, L203 discusses external API calls such as online hash services, which contradicts the local-only framing in the documentation and introduces a different operational model.
No suspicious patterns detected.
