The Agent Skills Directory

[COMMAND_EXECUTION] (MEDIUM): The skill executes shell scripts located outside its own directory at ~/.claude/skills/sentry-observability/scripts/detect_sentry.sh and list_alerts.sh.
Evidence: Found in SKILL.md under sections 'Error Tracking Check' and 'Alerting Check'. Execution of scripts from external skill paths can lead to unexpected behavior if those dependencies are modified or malicious.
[DATA_EXFILTRATION] (MEDIUM): The skill accesses the sensitive .env.local file to verify the presence of credentials like SENTRY_DSN and POSTHOG_KEY.
Evidence: grep -q "SENTRY_DSN|NEXT_PUBLIC_SENTRY_DSN" .env.local in SKILL.md.
Context: While this is a 'silent' check (grep -q) used for the primary purpose of an observability audit, direct access to environment files is a sensitive operation. Per the evaluation rules, this is downgraded from HIGH to MEDIUM as it is central to the skill's stated purpose.
[REMOTE_CODE_EXECUTION] (SAFE): No remote code execution patterns (e.g., curl piped to bash) were detected. All commands are executed against local files.
[PROMPT_INJECTION] (SAFE): No prompt injection or behavior override markers were detected in the instructions or metadata.

check-observability