The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill invokes uvx in SKILL.md and tools/fetch_conditions.py to download and execute the peakbagger-cli tool directly from the author's GitHub repository (github.com/dreamiurg/peakbagger-cli.git@v1.7.0).
[COMMAND_EXECUTION]:
The fetch_conditions.py script uses subprocess.run to execute shell commands for interacting with the PeakBagger CLI tool.
SKILL.md orchestrates various shell tasks, including running Python tools and dispatching specialized researcher agents that utilize the uv run python pattern.
The Report Reviewer agent has access to the Edit tool to modify generated report files.
[EXTERNAL_DOWNLOADS]:
The skill fetches real-time mountain data, weather forecasts, and avalanche information from several external sources including Open-Meteo, Mountain-Forecast.com, and the Northwest Avalanche Center (NWAC).
The tools/cloudscrape.py script uses the cloudscraper library to fetch HTML content from mountaineering websites (such as SummitPost and PeakBagger) to bypass bot protections.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it fetches and processes untrusted trip report content from various community-driven mountaineering websites.
Ingestion points: Researcher agents in SKILL.md (Phase 3B) ingest data from peakbagger.com, summitpost.org, wta.org, and alltrails.com.
Boundary markers: The agent prompts lack explicit instructions to ignore or delimit embedded commands within the fetched trip report text.
Capability inventory: The skill has the capability to write files (via the Report Writer agent) and execute shell commands (via Python tools).
Sanitization: Fetched web content is processed by LLM agents with minimal sanitization, relying on the agents to extract relevant information according to a JSON schema.

route-researcher