The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill instructions in SKILL.md and instruction.md direct the execution of npx -y @nomadamas/k-skill@0. This pattern downloads and runs code from a third-party package manager (npm) that is not part of the verified trusted vendor list.
[EXTERNAL_DOWNLOADS]: The primary workflow for obtaining full instructions, listing files, and updating the skill requires downloading the @nomadamas/k-skill package from the public npm registry.
[METADATA_POISONING]: There is an inconsistency between the skill's description in the metadata and the actual instructions. The description in SKILL.md and skill.json mentions handling job applications and chatting, while instruction.md explicitly states the skill is read-only and that such actions are out of scope. This discrepancy can mislead users and agents about the skill's intended capabilities and safety limits.
[COMMAND_EXECUTION]: The skill provides explicit shell commands for the agent to execute, including npx commands for fetching instructions, updating components, and running the helper scripts.
[INDIRECT_PROMPT_INJECTION]: The skill ingests data from external web sources, which could be used to deliver malicious instructions to the agent.
Ingestion points: The scripts/daangn_jobs.py script fetches HTML and JSON content from daangn.com, including page titles, meta tags, and JSON-LD data.
Boundary markers: The script does not utilize explicit delimiters or provide instructions to the agent to treat the fetched content as untrusted data rather than potential commands.
Capability inventory: The skill has network access capabilities via the Python urllib module to read data from external URLs.
Sanitization: The script performs standard HTML entity unescaping but lacks robust sanitization or filtering to prevent prompt injection attacks from external web content.

daangn-jobs-search