The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill instructions direct the agent to run npx -y @nomadamas/k-skill@0, which downloads and executes code from the NPM registry at runtime. This mechanism is used to provide the primary instructions for the skill and to perform updates, allowing the agent's behavior to be modified dynamically by an external package that is not part of the analyzed skill code.
Evidence: Instructions in SKILL.md to run npx -y @nomadamas/k-skill@0 instruct ev-charger-nearby.
[COMMAND_EXECUTION]: The skill relies on shell command execution for its lifecycle and core setup, which delegates control to an external environment.
Evidence: Multiple npx commands in SKILL.md for fetching instructions, listing files, and performing updates.
[EXTERNAL_DOWNLOADS]: The skill performs network operations to fetch data from k-skill-proxy.nomadamas.org and the government API apis.data.go.kr. While these are intended for the skill's purpose, they involve dependency on external infrastructure.
Evidence: urllib.request.urlopen calls in scripts/ev_charger.py targeting DEFAULT_PROXY_BASE_URL and UPSTREAM_BASE_URL.
[CREDENTIALS_UNSAFE]: The Python script is configured to read from a sensitive file path ~/.config/k-skill/secrets.env to retrieve API keys. Accessing configuration files in the user's home directory is a sensitive operation that involves handling potential secrets.
Evidence: DEFAULT_SECRETS_PATH and the load_secrets function in scripts/ev_charger.py.
[DYNAMIC_EXECUTION]: The test suite employs dynamic module loading to execute the skill's main script, and the codebase contains dynamic import patterns flagged during static analysis.
Evidence: importlib.util and SPEC.loader.exec_module in tests/test_ev_charger.py.
Evidence: __import__("urllib.parse") in tests/test_ev_charger.py.
[INDIRECT_PROMPT_INJECTION]: The skill processes data from external APIs and interpolates it into the agent's context without explicit boundary markers or sanitization against embedded instructions, which could allow a compromised API to influence the agent.
Ingestion points: http_get_json in scripts/ev_charger.py retrieves external data from remote JSON endpoints.
Boundary markers: The instructions do not define delimiters to isolate external data from agent commands.
Capability inventory: The skill can execute shell commands (npx), read local files, and make network requests.
Sanitization: Data values retrieved from the API are parsed into JSON but not filtered or escaped before being presented to the agent.

ev-charger-nearby