The Agent Skills Directory

[COMMAND_EXECUTION]: The skill's setup instructions in references/preflight-setup.md include commands that require elevated permissions. Specifically, it uses sudo apt install -y cpanminus to install system packages, which constitutes privilege escalation.
[REMOTE_CODE_EXECUTION]: The preflight setup script in references/preflight-setup.md performs remote code execution by piping the output of a network download directly into an interpreter: `curl -L https://cpanmin.us | mise exec perl -- perl
App::cpanminus. This is used to install the cpanminus` tool.
[EXTERNAL_DOWNLOADS]: The skill automates the installation of the Graph::Easy Perl module and the cpanminus installer from external repositories during its configuration phase as described in references/preflight-setup.md.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection due to its design of processing untrusted data from external files.
Ingestion points: The agent is instructed to read ADR files and extract content from specific <details> blocks as described in SKILL.md and references/adr-embedding.md.
Boundary markers: The workflow uses HTML tags (<details>, <summary>) and Markdown code blocks as delimiters, which are insufficient to prevent the agent from obeying instructions embedded within the diagram source code.
Capability inventory: The skill utilizes the Bash tool to execute shell commands and the graph-easy utility using data derived from the ingested Markdown files.
Sanitization: There is no evidence of validation or sanitization applied to the content extracted from files before it is passed to the execution environment.

adr-graph-easy-architect