The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it retrieves and processes untrusted content from Reddit, X (formerly Twitter), and general web search to generate instructions and prompts.
Ingestion points: The skill ingests data from external sources via scripts/lib/openai_reddit.py (Reddit), scripts/lib/xai_x.py (X), and Claude's internal web_search tool results.
Boundary markers: The skill uses simple Markdown headers (e.g., ### Reddit Threads) and separators (---) in its output to the agent, but it lacks specific instructions or robust delimiters to distinguish retrieved content from system instructions.
Capability inventory: The agent has the ability to execute shell commands via the skill's Python scripts and perform further network operations.
Sanitization: No evidence of sanitization, filtering, or escaping of the retrieved content was found in scripts/lib/normalize.py or the synthesis instructions in SKILL.md to prevent malicious payloads from influencing the LLM.
[COMMAND_EXECUTION]: The skill executes a local Python script to orchestrate the research and processing pipeline.
Evidence: SKILL.md contains the command python3 ~/.claude/skills/last30days/scripts/last30days.py "$ARGUMENTS" which runs the orchestrator.
[DATA_EXFILTRATION]: The skill performs network requests to external APIs to retrieve research data. These are directed at well-known services and are required for the skill's primary purpose.
Evidence: scripts/lib/http.py uses urllib.request to communicate with api.openai.com (OpenAI), api.x.ai (xAI), and www.reddit.com (Reddit).

last30days