skills/mukul975/anthropic-cybersecurity-skills/performing-sqlite-database-forensics/Gen Agent Trust Hub
performing-sqlite-database-forensics
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONMETADATA_POISONING
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes external SQLite database files (.db, -wal, -journal), which serve as untrusted data ingestion points. \n
- Ingestion points: The scripts/agent.py and scripts/process.py files read raw binary data and execute SQL queries on user-provided database files. \n
- Capability inventory: The scripts perform local file system operations, including reading database files and writing extracted binary pages and JSON reports to a user-specified output directory. \n
- Boundary markers: The instructions lack explicit boundary markers or warnings to the agent to treat data recovered from the databases (such as browser history or chat messages) as untrusted content that should not be interpreted as commands. \n
- Sanitization: While agent.py includes a regex check for table names (_SAFE_TABLE_RE), the actual content of the records is not sanitized before being included in the forensic report, creating a surface for indirect prompt injection if the agent is asked to analyze or summarize the findings. \n- [METADATA_POISONING]: There is a discrepancy between the author name in the SKILL.md frontmatter ('mahipal') and the copyright holder in the LICENSE file ('mukul975'). While likely a template oversight, inconsistent metadata can be used to obscure the origin of a skill.
Audit Metadata