The Agent Skills Directory

[EXTERNAL_DOWNLOADS] (MEDIUM): The README instructions direct users to install the skill via npx or git clone from the walterra/agent-tools repository. This source is not on the trusted organizations list, making it an unverifiable external dependency.\n
Evidence: README.md line 10: npx skills add walterra/agent-tools --skill es-ingest\n
Evidence: README.md line 13: git clone https://github.com/walterra/agent-tools.git\n- [COMMAND_EXECUTION] (MEDIUM): The skill utilizes a CLI script (ingest.js) that dynamically loads and executes JavaScript transformation files provided via the --transform flag. This allows for arbitrary code execution within the agent's environment.\n
Evidence: README.md line 26: --transform transform.js\n
Capability: The example scripts like examples/transform.js demonstrate that the tool executes user-provided logic on every document processed.\n- [PROMPT_INJECTION] (LOW): The skill ingests untrusted external data (CSV/JSON) which constitutes a surface for indirect prompt injection if the indexed data is later consumed by an LLM.\n
Ingestion points: scripts/ingest.js processes user-supplied files via the --file argument.\n
Boundary markers: Absent; there are no instructions provided to the agent to treat the data as untrusted or to use delimiters.\n
Capability inventory: The skill performs file system reads and network writes to Elasticsearch nodes.\n
Sanitization: Absent; the provided transformation examples focus on enrichment rather than security validation or sanitization of input data.

es-ingest