The Agent Skills Directory

[PROMPT_INJECTION]: No malicious override patterns, jailbreak attempts, or instructions to ignore safety guidelines were detected. The skill contains standard operational instructions and integrity rules for the agent.
[DATA_EXFILTRATION]: The skill does not perform network operations. All scripts use only the Python standard library for local file I/O. No hardcoded credentials or sensitive system path access (e.g., .ssh, .aws) were found.
[OBFUSCATION]: No Base64, zero-width characters, Unicode tags, or homoglyph attacks were detected in the codebase or metadata.
[REMOTE_CODE_EXECUTION]: No patterns for downloading or executing remote code were found. The skill does not use subprocesses, eval, or exec to run external commands or dynamic content.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data (VCF, BED, GTF, and FASTA files).
Ingestion points: Files are read via Path.read_text() and open() in scripts like audit_intervals.py and normalize_variant.py.
Boundary markers: None used; data is parsed based on tab-delimited or FASTA format expectations.
Capability inventory: Limited to local file read/write. No network access or shell command execution capabilities are present in the provided scripts.
Sanitization: Scripts validate field counts, ensure coordinates are integers, and check contig names against internal assembly tables.
[DYNAMIC_CONTEXT_INJECTION]: No use of the dynamic shell execution syntax (!command) was found in the SKILL.md file.
[PRIVILEGE_ESCALATION]: No use of sudo, chmod, or other privilege alteration commands was detected.
[PERSISTENCE]: The skill does not attempt to modify shell profiles, cron jobs, or system services for persistence.

genomic-coordinates