The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill fetches paper titles and abstracts from the external OpenAlex API and instructs the agent to analyze this content to write a detailed report. Maliciously crafted abstracts could attempt to influence the agent's behavior during the analysis phase.
Ingestion points: fetch_journal_papers.py retrieves untrusted text data (titles and abstracts) from api.openalex.org and saves it to /tmp/methods_scout/papers.json.
Boundary markers: There are no explicit instructions or delimiters provided to the agent to treat the fetched academic content as untrusted data or to ignore potential instructions embedded within them.
Capability inventory: The skill uses shell commands to run Python scripts (python3), performs network requests via urllib, and writes files using openpyxl and standard JSON libraries.
Sanitization: The classification script (heuristic_classify.py) uses regular expressions for categorization, but the agent reads the raw, unsanitized text in step 5 to write the final narrative report.
[COMMAND_EXECUTION]: The skill instructions in SKILL.md direct the agent to execute several local Python scripts and shell commands (mkdir, python3) to process data and generate reports. These commands use absolute paths specific to the author's local directory structure (/Users/songyiping/.claude/skills/...), which may fail or cause unexpected behavior on other systems.
[EXTERNAL_DOWNLOADS]: The script fetch_journal_papers.py connects to the OpenAlex API (api.openalex.org) to download academic metadata. While OpenAlex is a well-known and reputable academic service, this constitutes a network dependency for the skill's operation.

top-journals-methods-scout