The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute several shell commands for environment management and artifact retrieval.
It uses sf and gh CLI tools to interact with Salesforce orgs and GitHub repositories.
It performs cleanup operations like rm -rf ~/.sf/vscode-spans/ and process termination for port management using lsof and kill -9.
These actions are scoped to the development environment and are consistent with the skill's purpose.
[EXTERNAL_DOWNLOADS]: The skill uses the gh CLI to download test artifacts (logs, traces, and reports) from the vendor's repository (forcedotcom/salesforcedx-vscode). These downloads are performed from a trusted source related to the skill's author.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze untrusted data from external sources, which constitutes a vulnerability surface.
Ingestion points: The agent is instructed to read HTML reports (index.html), JSON test results, and span traces (.jsonl files) from the .e2e-artifacts/ directory.
Boundary markers: No specific delimiters or "ignore embedded instructions" warnings are defined for processing these artifacts.
Capability inventory: The agent has capabilities for shell command execution (gh, npm, sf, ffmpeg, rm, kill) and filesystem access.
Sanitization: There are no specific instructions for sanitizing or validating the content of the test artifacts before analysis.
An attacker who can influence test output (e.g., via a malicious Pull Request) could potentially embed instructions in the logs that the agent might follow during analysis.

playwright-e2e