The Agent Skills Directory

[DATA_EXFILTRATION]: The skill instructs the agent to search and extract data from broad local file paths, which could lead to unauthorized access to sensitive project data if not carefully scoped.
SKILL.md: The "Library-First Protocol" directs the agent to search D:\Projects\* to extract and adapt code from existing projects.
SKILL-meta.yaml: Contains an absolute local file path (C:\Users\17175\...) in the metadata field _migration.source_path, revealing internal directory structures and a local username.
[INDIRECT_PROMPT_INJECTION]: The skill defines a process for ingesting and embedding document corpora, which creates a vulnerability to indirect prompt injection if the ingested data contains malicious instructions.
Ingestion points: Document corpus processing in Phase 2 (SKILL.md, PROCESS.md).
Boundary markers: None identified; there are no instructions to use delimiters or ignore embedded commands within the processed documents.
Capability inventory: The skill uses the agentdb-vector-search library for database insertion and retrieval; no arbitrary command execution or system-level subprocess calls were identified.
Sanitization: There is no mention of sanitizing, escaping, or validating the document content before it is processed by the embedding model.

agentdb-semantic-vector-search