The Agent Skills Directory

[DATA_EXPOSURE]: The skill accesses sensitive local directories including ~/.copilot/session-state/, ~/.copilot/session-store.db, and platform-specific VS Code workspaceStorage paths to retrieve chat history and session metadata.
[COMMAND_EXECUTION]: The skill executes multiple shell commands to process data, including sqlite3 for database queries, and custom CLI tools qmd and obsidian-wiki for vault synchronization and indexing. It also uses environment variables (QMD_CLI) to determine which executable to run.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process external, untrusted data (Copilot chat transcripts) which may contain malicious instructions or prompt injection attempts.
Ingestion points: Reads from events.jsonl, session-store.db tables, VS Code transcript files, and memory artifacts (plan.md).
Boundary markers: The skill instructs the agent to use provenance markers like ^[inferred] and ^[ambiguous] to distinguish synthesized content and focuses on distilling knowledge rather than verbatim conversation copying.
Capability inventory: Performs file writes to the Obsidian vault and executes shell commands via sqlite3, qmd, and obsidian-wiki.
Sanitization: Includes explicit instructions to detect and skip secrets, API keys, and internal assistant reasoning fields (reasoningOpaque, reasoningText).
[DYNAMIC_EXECUTION]: The skill utilizes Python snippets for runtime decoding of Base64-encoded session identifiers found in VS Code storage paths.

copilot-history-ingest