The Agent Skills Directory

[PERSISTENCE]: The scripts/install.sh script modifies persistent agent directories by copying external skills and agents into ~/.claude/skills and ~/.claude/agents.
[DYNAMIC_EXECUTION]: The skill uses dynamic execution patterns in two places:
In references/report-pipeline.md, the agent is instructed to generate a Python script (generate_report.py) based on specific logic and then execute it using the python interpreter.
In references/deep-pipeline.md, the skill dynamically resolves the path to validate_json.py by probing several locations in the user's home directory and then executes the resolved path.
[EXTERNAL_DOWNLOADS]: The scripts/install.sh script clones code from a third-party GitHub repository (https://github.com/Weizhena/Deep-Research-skills) which is not identified as a known service.
[INDIRECT_PROMPT_INJECTION]: The skill has a significant indirect prompt injection surface:
Ingestion points: The skill ingests data from external web searches (via web-search-pipeline.md) and local YAML/JSON configuration files.
Boundary markers: No explicit delimiters or instructions are provided to the agent to disregard potentially malicious instructions embedded in the external search results.
Capability inventory: The skill utilizes python execution, file writing, and shell command execution (bash, grep).
Sanitization: There is no evidence of sanitization or filtering of the content retrieved from web searches before it is processed into reports or JSON data.
[COMMAND_EXECUTION]: The skill makes extensive use of shell commands in scripts/install.sh (git, cp, mkdir, pip) and references/deep-pipeline.md (ls, head, python) to manage its environment and execution.

deep-research