The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill connects to the OpenRouter API (openrouter.ai) to fetch current model pricing and retrieve usage statistics for the user's API key.
[COMMAND_EXECUTION]: The pipeline utilizes several standard command-line tools via the subprocess module to facilitate reporting:
gh (GitHub CLI) is used to list and view pull requests to link agent sessions to specific code changes.
git is invoked to determine repository remotes and metadata for local projects.
google-chrome is executed in headless mode to render generated HTML reports into PDF files.
sha256sum is used to maintain file integrity across mirrors during synchronization tasks.
[INDIRECT_PROMPT_INJECTION]: Because the skill processes raw agent transcripts and observation logs, it is exposed to untrusted data from previous model interactions.
Ingestion points: Processes local transcript files (~/.claude/projects/**/*.jsonl) and the Claude memory database (claude-mem.db).
Boundary markers: Implements parsing logic to identify distinct session boundaries and assistant/user turns.
Capability inventory: The skill uses Bash, Read, and Write tools to perform analysis and output reports.
Sanitization: Employs a specific redaction utility (behavior.py) that uses regular expressions to find and mask API keys (e.g., sk-or-), GitHub tokens, and other authentication headers in log excerpts to prevent accidental credential leakage in reports.

Agent Cost Report