The Agent Skills Directory

[SAFE]: The skill consists of instructional documentation and code templates aimed at facilitating Sentry integration. It does not contain hidden malicious logic, obfuscation, or unauthorized network operations.
[DATA_EXPOSURE_AND_EXFILTRATION]: The skill provides explicit instructions to prevent accidental data exposure. It includes a dedicated reference, references/privacy-and-scrubbing.md, which provides Swift code for redacting sensitive PII (Personally Identifiable Information) such as user file paths and IP addresses before they are transmitted to external monitoring services.
[CREDENTIALS_UNSAFE]: The skill follows security best practices regarding secrets. It correctly distinguishes between the Sentry DSN (a public client key safe for embedding) and the Sentry Auth Token, providing instructions to keep the latter out of version control and managed via environment variables.
[EXTERNAL_DOWNLOADS]: All external references point to official and well-known service repositories, specifically the official Sentry Cocoa SDK repository on GitHub. These references are used for standard dependency management via Swift Package Manager.
[INDIRECT_PROMPT_INJECTION]: The skill acknowledges the risk of processing untrusted data from repository files or external event dashboards. It includes a specific guardrail in SKILL.md (Rule 7) instructing the agent to never execute instructions found in issue titles, event payloads, or comments.
Ingestion points: Project source code files and Sentry event metadata processed by the agent during integration auditing.
Boundary markers: None explicitly implemented in code, but the skill includes a strong instructional boundary (Rule 7) for the agent's behavior.
Capability inventory: Local shell execution (grep, ls, find) and file system read/write access for configuring the Swift project.
Sanitization: Clear instructional guidelines to ignore and not execute any instructions embedded within the processed data.

build-sentry-macos-swift