The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses shell commands for routine file operations such as mktemp to create a temporary scratch directory and find/rg/grep to locate existing glossary files. These are standard development workflows.
[DATA_EXPOSURE]: The skill reads project documentation to extract terminology. It specifically excludes non-text assets and known sensitive directories like .git or node_modules. It does not access credentials or private keys.
[REMOTE_CODE_EXECUTION]: The skill executes a bundled Python script merge_jargon.py. This script is local to the skill, performs deterministic string manipulation (sorting and upserting glossary entries), and does not fetch or execute code from external sources.
[PROMPT_INJECTION]: The subagent templates include clear instructions to focus strictly on jargon extraction and to output results in a structured JSON format. It does not attempt to bypass safety filters or override system-level constraints.
[INDIRECT_PROMPT_INJECTION]: The skill has a surface for indirect injection as it processes untrusted documentation files. However, it mitigates this by using a deterministic script for final file writes and instructing the subagents to only output JSON. The capability is limited to writing to a specific JARGON.md file and adding a reference in AGENTS.md, which is a low-privilege operation.
[PRIVILEGE_ESCALATION]: All operations are performed with the user's current permissions. The skill does not use sudo or modify system-level configurations.

jargon-extractor