The Agent Skills Directory

[COMMAND_EXECUTION]: The skill employs a shell script, init-project.sh, to set up the _project/ directory structure. This script performs standard local filesystem tasks such as directory creation (mkdir) and template generation (cat) without requiring elevated privileges. These actions are within the expected scope of the skill's initialization process.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process external data including Jira exports, Slack history, and project documents, which presents a surface for indirect prompt injection attacks.
Ingestion points: External data is imported into the workspace during the 'initiate' and 'run-meeting' plays, referencing sources like project-brief.docx and meeting transcripts.
Boundary markers: The skill mitigates risks by requiring every data point to carry a provenance label (EV#, A#) and subjecting artifacts to an adversarial 'Judge (Solon)' audit persona.
Capability inventory: Capabilities include the execution of the init-project.sh script and regular read/write operations on markdown files within the localized project workspace.
Sanitization: Input integrity is maintained through verification against historical data in registers/lessons.md and evidence-based checklists in references/pm-discipline.md.
[EXTERNAL_DOWNLOADS]: The documentation provides standard installation instructions using npx skills add, a platform-native mechanism for incorporating agent skills. This is a routine operation and does not involve unauthorized remote code execution.

project-manager