The Agent Skills Directory

[COMMAND_EXECUTION]: The skill requires the agent to execute shell-based curl commands to communicate with the opencode-bridge:8000 service. This includes uploading local files and downloading generated assets to temporary storage locations.
[DATA_EXFILTRATION]: The skill is configured to read local files and transmit them to a network endpoint. It also embeds user-specific identifiers such as open_id or chat_id directly into the URL paths for these network requests, which may lead to the exposure of user metadata to the internal bridge service.
[INDIRECT_PROMPT_INJECTION]: The skill's core functionality involves ingesting and analyzing potentially untrusted file types (Excel, CSV, PDF, Word) provided by users. 1. Ingestion points: The skill uploads user files to the processing bridge and interprets the resulting textual and file-based outputs. 2. Boundary markers: There are no explicit instructions or delimiters provided to the agent to treat the data returned from the bridge as potentially untrusted or to ignore instructions embedded within those files. 3. Capability inventory: The agent possesses network and file system access via curl, along with the ability to create and modify documents through feishu_doc and feishu_media tools. 4. Sanitization: The instructions do not mandate the sanitization of content extracted from processed files before it is used to generate final summaries or reports.

opencode-writer