The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill and its included framework scripts ingest data from local repository files which could be manipulated to influence agent behavior.
Ingestion points: SKILL.md (Step EJ1) reads epic markdown files at EPIC_FILE_PATH. references/defer-mutation.js parses the references/tracker-access-record.md file to determine valid mutation kinds and consequences.
Boundary markers: Absent; the skill and scripts parse YAML frontmatter or markdown tables from these files without explicit sanitization or instructions to ignore embedded instructions in the ingested data.
Capability inventory: The skill invokes Atlassian MCP tools (getJiraIssue) and executes Node.js scripts (sync-jira-epic.js). It also performs file system writes to update markdown frontmatter.
Sanitization: Absent; extracted values (such as jira_key, title, and roster kinds) are used directly as parameters for tools and command-line execution.
[DYNAMIC_EXECUTION]: Internal helper scripts utilize dynamic code execution for environment resolution and configuration parsing.
references/read-config.sh executes Python code strings via python -c to parse YAML configurations when the pyyaml library is available.
references/defer-mutation.js uses child_process.spawnSync to execute bash commands for platform detection and access mode resolution.
[COMMAND_EXECUTION]: The skill executes a bundled script in Step EJ4 by running node .agents/skills/sync-jira-epic/scripts/sync-jira-epic.js --file "$EPIC_FILE_PATH" to synchronize data between the local file and Jira.

ensure-epic-jira-issue