The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted data from local directories, creating a significant attack surface for indirect instructions.
Ingestion points: The skill explicitly targets files in the incoming/ directory for 'Data Intake Review' and .eml files in the eml/ directory for 'Communications Triage'.
Boundary markers: The prompt templates (e.g., intake-review.md, email-triage.md) do not include delimiters or instructions to ignore embedded commands within the source data.
Capability inventory: The skill demonstrates workflows for reading files, executing CLI commands via openclaw, and writing report artifacts to the file system.
Sanitization: There is no evidence of content sanitization, filtering, or validation before the data is processed by the AI assistant.
[EXTERNAL_DOWNLOADS]: The skill instructs users to download external code from a non-verified personal repository during installation.
Evidence: git clone https://github.com/dandenney/webinars-build-your-own-executive-assistant-with-openclaw.git in the Installation section.
[COMMAND_EXECUTION]: The skill provides patterns for executing shell commands and scripts on the local system.
Evidence: Documentation includes a shell script wrapper (generate-daily-log.sh) that invokes the openclaw CLI tool with arguments derived from local file content.
[PERSISTENCE]: The skill encourages the use of system scheduling tools to maintain automated workflows across sessions.
Evidence: The 'Automation example' section provides crontab configurations to periodically execute shell scripts that process local data.

openclaw-executive-assistant-webinars