The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process data from various untrusted external sources, creating a surface for indirect prompt injection.
Ingestion points: The skill reads data from CSV and TSV files (pd.read_csv in data-manipulation-recipes.md and enrichment-patterns.md), parses phylogenetic tree files in Newick format (re.findall in species-reconciliation.md), and processes output from AWS S3 listing commands.
Boundary markers: The instructions do not specify the use of clear delimiters or instructions to ignore embedded commands within the processed data.
Capability inventory: The skill uses the Bash tool (listed in allowed-tools in SKILL.md) and Python's subprocess.run to interact with the system.
Sanitization: There is no evidence of sanitization or validation of the content of the external data strings before they are processed or used to generate analysis interpretations.
[COMMAND_EXECUTION]: The skill includes Python patterns that execute shell commands using the subprocess module.
Evidence: enrichment-patterns.md contains code patterns using subprocess.run(['aws', 's3', 'ls', path, ...]) to validate the existence of data on AWS S3.
[EXTERNAL_DOWNLOADS]: The skill interacts with external infrastructure to verify and potentially fetch data from a well-known service.
Evidence: enrichment-patterns.md describes a workflow for inferring and validating paths on AWS S3 using the aws CLI.

data-analysis-patterns