docs: AGENTS.md in first person throughout #51
No reviewers
Labels
No labels
Agent
Compat/Breaking
Kind/Bug
Kind/Documentation
Kind/Enhancement
Kind/Feature
Kind/Security
Kind/Testing
Priority
Critical
Priority
High
Priority
Low
Priority
Medium
Reviewed
Confirmed
Reviewed
Duplicate
Reviewed
Invalid
Reviewed
Won't Fix
Status
Abandoned
Status
Blocked
Status
Need More Info
No milestone
No assignees
1 participant
Notifications
Due date
No due date set.
Dependencies
No dependencies set
Reference
joshtronic/igor!51
Loading…
Reference in a new issue
No description provided.
Delete branch "docs/agents-first-person"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
The system prompt loads AGENTS.md as part of Igor's identity-and-
rules layer. Mixing "you are X" (external authority addressing the
agent) with "I am Y" (self-description) creates cognitive
inconsistency: rules are imposed at him, but other parts of the
same prompt are spoken in his voice.
Converting AGENTS.md to first person makes the whole system-prompt
layer read like Igor's own operating manual that he wrote, not
rules being declaimed at him. Aligns with the broader autonomy
direction (identity earned, dealer's-choice journaling, gatekeeping
via PR review instead of prescription).
Operational hard rules ("Never commit to main", "Don't push") work
in first person too -- imperatives don't need a second-person
subject.
Three "you" references remain inside quoted strings -- those are
literal text the harness sends in user messages, so they describe
what the message will say to me. Keeping them quoted preserves
accuracy.
The tick.sh-generated user messages (which address Igor in second
person) are intentionally NOT changed -- those are external context
handed to him per-tick, not self-description.
Co-Authored-By: Claude Opus 4.7 noreply@anthropic.com
The system prompt loads AGENTS.md as part of Igor's identity-and- rules layer. Mixing "you are X" (external authority addressing the agent) with "I am Y" (self-description) creates cognitive inconsistency: rules are imposed at him, but other parts of the same prompt are spoken in his voice. Converting AGENTS.md to first person makes the whole system-prompt layer read like Igor's own operating manual that he wrote, not rules being declaimed at him. Aligns with the broader autonomy direction (identity earned, dealer's-choice journaling, gatekeeping via PR review instead of prescription). Operational hard rules ("Never commit to main", "Don't push") work in first person too -- imperatives don't need a second-person subject. Three "you" references remain inside quoted strings -- those are literal text the harness sends in user messages, so they describe what the message will say to me. Keeping them quoted preserves accuracy. The tick.sh-generated user messages (which address Igor in second person) are intentionally NOT changed -- those are external context handed to him per-tick, not self-description. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>