The Agent Skills Directory

[COMMAND_EXECUTION]: The SKILL.md workflow explicitly instructs the agent to execute a local Python script using uv run scripts/refresh_sources.py. This script is used to synchronize the skill's reference materials with upstream sources.
[EXTERNAL_DOWNLOADS]: The refresh_sources.py script performs network requests to fetch documentation from docs.deckset.com and example presentations from api.github.com. These downloads target the official product documentation and well-known services.
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection due to its reliance on external data sources.
Ingestion points: The agent reads content from references/deckset-markdownDocumentation.html and references/examples/*.md which are populated via external network requests.
Boundary markers: There are no explicit instructions in the SKILL.md to treat the content of these reference files as untrusted or to ignore embedded instructions within them.
Capability inventory: The skill allows the execution of a local Python script via uv run and the generation of markdown files that are influenced by the fetched data.
Sanitization: While the Python script cleans HTML tags and unescapes characters, it does not perform any validation or filtering to prevent the agent from obeying malicious instructions embedded in the external markdown files.
[CREDENTIALS_UNSAFE]: The Python script checks for the existence of GH_TOKEN or GITHUB_TOKEN in the environment to authenticate with the GitHub API. This is a standard and safe method for handling API access, as no secrets are hardcoded in the skill itself.

deckset