How Far to Narrow an Agent's Choices in a 30-Category Wallpaper Classification Pipeline
Asking an agent to pick one of 30 categories per image means re-running every image the moment a definition changes. Here is the reasoning and the implementation behind switching to closed-vocabulary tags plus a deterministic rule mapping.
A Typo in agent.md Quietly Widened My Permissions — Writing a Strict Frontmatter Lint
Misspelled keys in agent.md frontmatter do not raise errors. They fall back to defaults, and for permission fields that fallback points the wrong way. Here is the failure I hit, the lint I wrote to catch it, and what the measurements showed.
After Compose-First: Choosing Which View Screens to Migrate, Ranked by Churn Instead of Count
Google has declared Android development Compose-first. Here is how I rank View-based screens for migration using git history rather than screen counts, with the scoring script I actually ran and the three places partial migration quietly duplicates state.
One Space in a Path, and Nine Commands Reported Success While Counting the Wrong Place
A single space in a workspace name sends agent-written commands somewhere else, quietly. Measurements across eleven unquoted-path forms, and the entry-point script that closes the boundary in one cd.
Android 17's Local Network Permission: Inventory Every LAN Call Site Before You Hand the Fix to an Agent
Apps targeting the new API level need explicit permission to reach the local network. A grep for private IPs found 2 of my 11 LAN paths. Here is the detector that found all 11, and how I turned its output into eleven checkable tasks for an agent.
Three Quarters of My Reference Notes Never Reached the Agent: Measuring What head Cuts Away
I fed reference notes to a scheduled agent with cat and head, and the lines that mattered were quietly cut. Here is the measurement, and how I replaced a line count with a section-level contract.
When to Hand Your Agent the Next Instruction: Waiting, Interrupting, and Queuing, Measured
Antigravity v2.3.0 added message queuing and Send Now. I measured waiting, interrupting, and queuing against the same yardstick, found that 41% of queued instructions arrive stale, and cut rework from 22% to 9% with a twenty-line stamp.
The File Is Right There in ls, and Your Agent Still Can't Open It
The agent says the file does not exist. Your terminal says it does. After three days of blaming cloud sync, the answer turned out to be that one voiced consonant mark was never a single character. Detection script and a three-layer gate included.
Passing API Keys to Agents Safely: Runtime Env Injection and Log Redaction
How to hand secrets to an Antigravity agent without leaving them in your repo or your logs, using runtime environment injection and output masking.
What to Delegate to an Antigravity Agent and What to Keep by Hand, After Two Weeks
After two weeks of handing my daily solo-dev tasks to Antigravity agents, a clear line emerged between the work I was glad to delegate and the work I had to pull back. A retrospective with the operational log.
The Scarier Permission Was Reach, Not Write: Locking an Agent's Outbound Traffic to an Allowlist
When I hand real work to an agent, the thing I guard most isn't file edits, it's where it connects. Here is an 80-line deny-by-default egress gate, plus what 21 nights of unattended runs revealed about the traffic I never saw.
Clock Design for Agents: Separating Monotonic and Wall Time to Protect Long Runs
In scheduled and overnight agent runs, computing timeouts, retry backoff, and rate-limit windows from the wall clock breaks in quiet, hard-to-debug ways. This piece walks through a small Clock abstraction that splits monotonic from wall time, with a decision table and testable code.