The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes data from external advertising platforms (Google Ads, Meta, LinkedIn, TikTok) and local profile/guideline files. While this data is typically numeric or structural, if these sources were compromised to include natural language instructions, they could potentially influence the agent's behavior during analysis.
Ingestion points: Advertising platform spend/campaign metrics, profile.json, and brand guidelines.
Boundary markers: None identified in the instruction set.
Capability inventory: Executes local Python scripts (ad-budget-pacer.py, budget-optimizer.py, performance-monitor.py) and writes spend snapshots to the filesystem.
Sanitization: No explicit sanitization or filtering logic is defined for the ingested data before it is passed to scripts.
[DYNAMIC_EXECUTION]: The skill generates shell commands to execute local Python scripts using variable interpolation from user input and brand profiles.
Evidence: The process steps include commands like python "${CLAUDE_PLUGIN_ROOT}/scripts/ad-budget-pacer.py" --budget {total} --period-days {N} --days-elapsed {N} --spend-to-date {amount}.
Context: These scripts are part of the skill's own internal distribution (${CLAUDE_PLUGIN_ROOT}/scripts/) and are used for pacing and optimization calculations.
[COMMAND_EXECUTION]: The skill uses command-line operations to perform its core functions, including budget pacing analysis, optimization, and snapshot persistence.
Evidence: Multiple steps in the Process section invoke Python scripts via the shell to handle data processing and storage tasks.

budget-tracker