The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from external URLs and local files, creating a potential vector for indirect prompt injection.\n
Ingestion points: The skill reads file paths provided by the user and fetches content from external URLs using the Jina AI reader service.\n
Boundary markers: The instructions do not specify the use of strong delimiters to isolate untrusted content from the agent's instructions.\n
Capability inventory: The skill can execute shell commands (curl), run a local Python script (fix_punctuation.py), and perform file read/write/delete operations in a temporary workspace.\n
Sanitization: There is no evidence of specific sanitization or filtering of the ingested content to prevent embedded instructions from influencing the agent's behavior.\n- [DYNAMIC_EXECUTION]: The skill executes a bundled Python script (scripts/fix_punctuation.py) to process the draft translation. While the script performs legitimate punctuation correction, the execution of local scripts at runtime is a dynamic code execution pattern.\n- [COMMAND_EXECUTION]: The skill invokes the curl command to fetch web content. This allows the skill to perform network operations based on user-supplied URLs.\n- [EXTERNAL_DOWNLOADS]: The skill fetches content from the Jina AI service (r.jina.ai). This is a well-known service used to facilitate the extraction of clean text from web pages for processing by AI agents.

translate-polisher