The Agent Skills Directory

[COMMAND_EXECUTION]: The shell script scripts/run-komaci.sh uses the KOMACI_ESLINT_BIN environment variable to determine the path to the ESLint binary. Because this variable is expanded without quotes during execution, an attacker with control over the environment could inject arbitrary commands or execute untrusted programs.
[EXTERNAL_DOWNLOADS]: The scripts/run-komaci.sh script automatically executes npm install to retrieve the @salesforce/eslint-plugin-lwc-graph-analyzer and eslint packages if they are not already present in the script directory. This process downloads code from the public NPM registry at runtime. These packages belong to a well-known technology organization.
[PROMPT_INJECTION]: The skill processes external JavaScript, TypeScript, and HTML files as part of its review workflow, creating an indirect prompt injection attack surface.
Ingestion points: Source code files for Lightning Web Components are read from the local filesystem and analyzed by the agent (SKILL.md, Steps 3-5).
Boundary markers: The skill lacks explicit instructions or markers to isolate the untrusted component code from the agent's internal reasoning or to prevent the agent from following instructions embedded within the analyzed code.
Capability inventory: The skill can execute shell scripts (scripts/run-komaci.sh) and has instructions to automatically apply code fixes (SKILL.md, Step 7), which involves modifying the component's source files.
Sanitization: No sanitization or validation of the input source code is performed before it is used for analysis or before the agent suggests/applies modifications.

reviewing-lwc-mobile-offline