The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process data from external, untrusted sources including web pages and GitHub issues.
Ingestion points: External content is retrieved using WebFetch for URLs and gh issue view for GitHub issues (SKILL.md, Step 5).
Boundary markers: The skill uses an AskUserQuestion confirmation gate (Yes/No) and is instructed to restate the idea in its own words before performing writes, which helps mitigate accidental instruction following.
Capability inventory: The agent has the ability to write to journal/ideas/<slug>.md and update journal/JOURNAL.md. It also has access to the local shell and GitHub CLI.
Sanitization: No programmatic sanitization of the fetched text is described; the skill relies on user review during the confirmation step.
[COMMAND_EXECUTION]: The skill executes local commands to interact with the project environment and version control systems.
Evidence: SKILL.md contains instructions to execute python -m skore_skills status, gh auth status, and gh issue view to retrieve project state and external data.
[EXTERNAL_DOWNLOADS]: The skill retrieves content from external network locations.
Evidence: Step 5 in SKILL.md explicitly instructs the agent to use WebFetch on user-provided URLs and documentation pages.

shape-user-idea