The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill requires executing code from a remote registry to initialize and update its instructions.
SKILL.md directs the agent or user to run npx -y @nomadamas/k-skill@0 instruct sh-notice-search as a mandatory setup step.
SKILL.md includes an update command npx -y @nomadamas/k-skill@0 update that fetches and executes code from the @nomadamas npm scope.
[COMMAND_EXECUTION]: The skill provides instructions for executing local scripts to perform its core functions.
instruction.md contains examples for running node packages/sh-notice-search/src/cli.js with various arguments.
[EXTERNAL_DOWNLOADS]: The skill retrieves data from the official Seoul Housing Corporation public website.
It makes automated requests to www.i-sh.co.kr to fetch notice listings and details.
[INDIRECT_PROMPT_INJECTION]: The skill ingests and interprets HTML from an external source, presenting a risk of indirect instruction injection.
Ingestion points: Notice lists and detail pages are fetched from www.i-sh.co.kr (referenced in instruction.md).
Boundary markers: No explicit delimiters are defined to separate ingested data from agent instructions.
Capability inventory: The skill environment allows script execution and network access.
Sanitization: The instructions do not describe any sanitization of the parsed HTML before it is presented to the agent.
[METADATA_POISONING]: A discrepancy exists between the skill's metadata and its internal logic.
skill.json declares the action:submission profile.
However, instruction.md and SKILL.md contain strict rules against automated submissions, creating conflicting signals regarding the skill's intended behavior.

sh-notice-search