The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes a bash subshell to execute a local script for resolving directory paths.
Evidence: The instruction in SKILL.md to search $(bash ~/.claude/coordination/resolve-coord-dir.sh)/handoffs/ involves the execution of a script from a hidden user directory to compute search paths at runtime.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it loads and follows instructions from external handoff files without sanitization.
Ingestion points: SKILL.md (Step 1) specifies reading files from docs/summaries/, .claude-sessions/, and worktree directories.
Boundary markers: None present; the skill does not use delimiters or warnings to prevent the agent from obeying instructions embedded within the handoff files.
Capability inventory: The skill allows the agent to navigate the file system (cd) and execute repository management tools (git status, git branch).
Sanitization: None present; the skill does not validate the content of the markdown files before processing the 'Remaining Work' and other instructional sections.

session-resume