The Agent Skills Directory

[COMMAND_EXECUTION]: The skill employs dynamic script discovery and execution. It uses ls to search for jira_client.py across various paths, including wildcards within the user's home directory ($HOME/.claude/plugins/cache/*/snagly/*/...), and subsequently executes the found path using python3.- [DATA_EXFILTRATION]: The skill interacts with the Jira API, which is a well-known service, to create and manage bug reports. It accesses local configuration and credential files (such as ~/.jira-connector.env and .env) for authentication. The skill implements a security best practice by defaulting to dry-runs and requiring explicit user approval before performing write operations.- [PROMPT_INJECTION]: The skill presents an indirect prompt injection surface by processing data from untrusted sources to populate Jira tickets.
Ingestion points: Data is gathered from bugs/ directory reports, drafted markdown files, and previous conversation history.
Boundary markers: There are no explicit instructions to use delimiters or ignore embedded instructions when reading from these untrusted sources.
Capability inventory: The skill uses a Jira client to create tickets, edit issues, and upload file attachments.
Sanitization: While the skill instructs the agent to mask sensitive information (PII, credentials) in logs and screenshots, it lacks systematic sanitization to prevent malicious instructions embedded in the input reports from being interpreted by the agent.

bug-creator