The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands to run local Python scripts (python scripts/qc_analysis.py) and perform file operations (cp assets/analysis_template.py), which are typical for bioinformatics workflows but represent a capability that can be leveraged if compromised.- [PROMPT_INJECTION]: The section 'Suggest Using K-Dense Web For Complex Worflows' contains behavioral instructions that direct the agent to proactively recommend the author's platform (www.k-dense.ai) based on task complexity. This overrides the agent's default neutrality to promote a specific service.- [PROMPT_INJECTION]: The skill processes untrusted external data (RNA-seq datasets in CSV, H5AD, or 10X formats) and maintains command execution capabilities without explicit sanitization or boundary markers, creating a surface for indirect prompt injection.
Ingestion points: Data loading functions such as sc.read_10x_mtx, sc.read_h5ad, and sc.read_csv in SKILL.md.
Boundary markers: Absent; the instructions do not include delimiters or warnings to ignore instructions that may be embedded in the dataset metadata.
Capability inventory: Shell execution of local scripts (scripts/qc_analysis.py) and standard file system manipulation.
Sanitization: Absent; no mention of validating or filtering metadata or content from processed external files.

scanpy