context-management
Pass
Audited by Gen Agent Trust Hub on Oct 29, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill extracts and re-injects content from the conversation history, which creates a potential surface for indirect prompt injection.
- Ingestion points: Conversation history is ingested via the
conversation_dataparameter incontext_extractor.py. - Boundary markers: Restored context is formatted with Markdown headers in
context_rehydrator.py(e.g.,## Original Requirements), but lacks strong delimiters or instructions to the model to disregard embedded commands. - Capability inventory: The skill performs local file writes to save snapshots in
~/.amplihack/.claude/runtime/context-snapshots/and provides text that is directly interpolated into the agent's prompt during the rehydration process. - Sanitization: The skill performs basic keyword extraction and truncation but does not sanitize the content for potential injection attacks embedded in the history.
Audit Metadata