The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/readwise_to_nlm.py interacts with the local system by executing the nlm utility. The command is invoked using a list of arguments via subprocess.run, which effectively prevents shell injection vulnerabilities.
[DATA_EXPOSURE]: The skill scripts access a Readwise API token stored at a specific filesystem path (/var/folders/01/wzs3mqmn3jx2b81f0dcq9w8h0000gq/T/agenix/readwise-token) for authentication. While this is a sensitive operation, it is a functional requirement and uses a path consistent with specific secret management tools.
[INDIRECT_PROMPT_INJECTION]: The skill ingests and processes content (articles, highlights, HTML) from the external Readwise API, creating a risk that malicious instructions embedded in saved documents could influence the agent's behavior.
Ingestion points: scripts/readwise_to_nlm.py and scripts/readwise_prune.py fetch data from readwise.io/api/v3/list/.
Boundary markers: The skill documentation and scripts do not implement explicit delimiters or instructions to treat the ingested data as untrusted text.
Capability inventory: The skill has the capability to perform network requests (requests) and execute local CLI commands (nlm).
Sanitization: Although scripts/readwise_to_nlm.py uses html2text for Markdown conversion, it lacks specific filters or sanitization logic to detect or strip potential prompt injection payloads from the retrieved text content.

readwise