get-yearly-fx-rate

Pass

Audited by Gen Agent Trust Hub on Sep 15, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill fetches annual average exchange rate data from the official IRS website (irs.gov) using the urllib.request module. This network operation targets a well-known and trusted government service and is central to the skill's primary function.
  • [COMMAND_EXECUTION]: The skill executes local Python scripts (get_yearly_fx_rate.py and _workpaper.py) to parse data and generate documentation. These scripts perform standard file operations, such as directory creation and the copying of proof artifacts (e.g., HTML snapshots, PDFs, or screenshots), to build a structured audit workpaper. The code uses standard library components and includes self-test functionality to ensure reliability.
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from the IRS website or locally supplied HTML snapshots. To mitigate risks associated with this attack surface, the skill implements the following:
  • Ingestion Points: scripts/get_yearly_fx_rate.py reads HTML content from a hardcoded URL or a user-specified path.
  • Boundary Markers: Detailed instructions in SKILL.md and references/source-policy.md enforce a strict retrieval and documentation workflow.
  • Capability Inventory: Scripts have capabilities for network fetching (urlopen), file writing, and file copying (shutil.copy2).
  • Sanitization: The skill employs a dedicated TableParser (subclass of HTMLParser) to extract data, along with text cleaning utilities and PDF-specific character escaping and normalization to ensure data is handled safely before being presented to the user or rendered into files.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 15, 2026, 08:16 AM
Security Audit — agent-trust-hub — get-yearly-fx-rate