The Agent Skills Directory

[PROMPT_INJECTION]: The skill presents a surface for indirect prompt injection as it is designed to ingest and process external analytics reports and user briefs to influence its output.
Ingestion points: The agent reads data from brief.md and various text snapshots in the data/ directory.
Boundary markers: The instructions rely on standard Markdown formatting but do not include specific delimiters or instructions to ignore potential commands embedded within the external data snapshots.
Capability inventory: The skill possesses file-writing capabilities (results.tsv, final_variants.md) and shell execution capabilities (npx, bash scripts).
Sanitization: The skill processes external data content directly into the variant generation loop without explicit sanitization or validation steps.
[EXTERNAL_DOWNLOADS]: The skill fetches a versioned CLI tool from the vendor's official package namespace.
Evidence: SKILL.md and references/brief-template.md use npx --yes @agent-analytics/cli@0.5.31 to perform analytics operations. This is a vendor-owned resource pinned to a specific version.
[COMMAND_EXECUTION]: The skill executes local bash scripts and system binaries to initialize the environment and process data.
Evidence: The skill invokes scripts like scripts/init_autoresearch_run.sh and scripts/collect_agent_analytics_snapshot.sh to manage workspace setup and data collection. It also uses perl for basic text processing on command output.

agent-analytics-autoresearch