The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze untrusted external data, specifically entire codebases, git histories, and scattered notes (Phase 0, Phase 1, Phase 1-H).
Ingestion points: Phase 0 ("codebase, an invention disclosure, scattered notes") and Phase 1-H ("git repository") ingest data directly into the agent's context.
Boundary markers: No specific delimiters or instructions to ignore embedded malicious content within the source code are mentioned.
Capability inventory: The skill uses Bash, Read, Write, and specialized patent search tools to process the ingested data.
Sanitization: No sanitization or filtering of the ingested source code is described before it is used to generate claims or search queries.
[COMMAND_EXECUTION]: The skill makes extensive use of the Bash tool to execute complex shell commands and pipes for git history analysis, such as git rev-list --all | git diff-tree --stdin -p -r | git patch-id --stable. This provides a significant execution surface when operating on untrusted repositories.
[DATA_EXFILTRATION]: The skill reads potentially sensitive local information (proprietary source code and invention disclosures) and subsequently performs network-based search operations through tools like search_patents_bigquery and external "web-research agents." This creates a channel where fragments of sensitive or secret information contained within the analyzed code could be included in search queries sent to external services.

patent-application-creator