weekly-digests

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (medium risk: 0.65). The skill fetches a full claude-mem timeline via a runtime HTTP call (curl ... /api/context/inject?project=...&full=true) and then reads/splits that resulting markdown into per-week files that are injected into each subagent prompt as {ABSOLUTE_PATH_TO_WEEK_FILE}; that timeline content can include outsider-authored free text (e.g., comments/issues/PRs/wiki/discussion entries) authored by non-operating users.

MEDIUM W021: Hidden or invisible Unicode characters detected (potential obfuscation or prompt injection).

• Hidden Unicode characters detected (1 type(s) found)