The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill's core functionality involves scanning and ingesting data from various sources to populate a persistent evidence ledger.
Ingestion points: The /evidence onboard command and the commitment-sweep.ts script scan local project files (e.g., .agents/context/*.md, Client-Intake/, README.md) and fetch data from GitHub issues using the gh CLI tool.
Boundary markers: While the skill uses Markdown formatting and horizontal rules to separate entries, it does not implement explicit delimiters or instructions to ignore potential commands embedded within the ingested data.
Capability inventory: The agent has access to sensitive tools including bash, write_to_file, and replace_file_content, which could be exploited if malicious instructions are successfully injected via project files or issue descriptions.
Sanitization: There is no evidence of sanitization or escaping of the ingested text before it is stored in the ledger or used in project briefings.
[COMMAND_EXECUTION]: The skill uses the gh (GitHub CLI) tool within its commitment-sweep.ts script to query open issues for commitments. It also encourages the use of shell commands (via the [RAW] confidence tier) to generate verification receipts, which is consistent with its stated purpose of evidence tracking.

evidence-ledger