The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted external data from URLs and PDFs using the web_extract tool. This content is then summarized and processed by the agent to create or update wiki pages, creating a potential vector for instructions embedded in external sources to influence agent behavior.
Ingestion points: Raw source material is fetched from external URLs and stored in the raw/ directory via web_extract as described in SKILL.md.
Boundary markers: The instructions do not specify the use of delimiters or boundary markers to isolate source content from agent instructions during processing.
Capability inventory: The skill possesses capabilities for read_file, search_files, web_extract, execute_code, and shell command execution for synchronization tasks.
Sanitization: There are no explicit sanitization or filtering steps mentioned for content retrieved from external sources.
[PERSISTENCE]: The skill includes instructions for setting up a systemd user service (obsidian-wiki-sync.service) to ensure that a synchronization process runs continuously in the background.
[DYNAMIC_EXECUTION]: The skill employs execute_code to run Python scripts generated at runtime for linting the wiki, such as identifying orphan pages by scanning for [[wikilinks]] across the filesystem.
[EXTERNAL_DOWNLOADS]: The documentation directs users to install the obsidian-headless package via npm and references the llm-wiki-compiler tool from GitHub, both of which originate from sources outside the trusted vendor list.
[PRIVILEGE_ESCALATION]: Setup instructions involve the use of sudo loginctl enable-linger $USER, a command that modifies system behavior to allow user processes to persist after logout.
[CREDENTIALS_UNSAFE]: The skill provides an example of a CLI login command (ob login --email <email> --password '<password>') that encourages passing sensitive credentials as plain-text arguments, potentially exposing them in shell history.
[COMMAND_EXECUTION]: The skill involves the execution of various shell commands for package management, service control, and synchronization workflows.

llm-wiki