Bug Tracking for Stealth Games

Stealth games live on the tension of avoiding detection, and that tension depends entirely on AI perception working correctly and predictably. When detection misfires, a guard spotting the player through a wall, an enemy ignoring the player standing in plain sight, an alert state that triggers wrong, the core experience breaks, because the player can no longer trust the rules of sneaking. These bugs depend on the AI perception state, the line-of-sight calculation, and the player position, none of which the player can see. Tracking stealth bugs means capturing that AI-and-sightline context to reproduce a detection that went wrong.

Detection is the core, and it is fragile

The defining mechanic of a stealth game is detection: the AI perceives the player through sight, sound, and other senses, and the player avoids that perception. This makes AI perception the load-bearing system, and its bugs especially damaging, because when detection misfires, the player loses trust in the rules they are supposed to play by. A guard who spots the player through a wall, or fails to see them in the open, breaks the fundamental contract of the genre.

This fragility means detection bugs are core gameplay failures, not minor glitches. The player needs detection to be consistent and fair to plan their stealth, and an unpredictable or wrong detection ruins the experience. These bugs are also hard to report, since the player sees the symptom, spotted unfairly, ignored impossibly, but not the AI perception state, line of sight, or sound calculation that produced it. Capturing that hidden state is what makes stealth detection bugs diagnosable.

Capture the AI perception state

The core context for a stealth bug is the AI perception state at the moment of the bug: the detection level of the relevant AI, whether they were unaware, suspicious, or alerted, and what was driving that state. When a player reports being spotted unfairly or ignored impossibly, capturing the AI detection state reveals what the AI actually perceived and why it was in that state.

Capture which senses contributed, sight, sound, other, and their values, since AI perception usually combines multiple inputs into a detection level, and a bug often hides in how these combine, a sound that should not have alerted, a sight detection that fired through cover. The AI perception state, broken down by sense, shows you what the AI was perceiving when it made the detection decision that the player experienced as wrong, which is the starting point for finding the perception bug.

Capture the line of sight and position

Line of sight is central to stealth detection, and many stealth bugs are sightline bugs: the AI seeing through a wall or object it should not see through, or failing to see the player in clear view. Capture the line-of-sight calculation, what the AI sightline check determined, along with the positions of the AI and the player and the geometry between them, since the bug is about whether the AI could actually see the player.

A spotted through a wall report becomes diagnosable when you can see that the line-of-sight check returned a hit it should not have, perhaps because of a gap in the occluding geometry or a flaw in the sightline logic. The AI and player positions and the intervening geometry let you reconstruct the sightline and see whether the detection was geometrically justified. Capturing the line-of-sight context and the positions turns a sightline complaint into a visible geometric and logical question you can answer.

Capture sound and alert propagation

Beyond sight, stealth games model sound, the noise the player makes and how it propagates to alert AI, and alert states that spread among enemies. Sound and alert bugs are common: a noise that alerts an AI it should not reach, an alert state that propagates wrong among guards, a detection that lingers or clears incorrectly. Capture the sound state and alert propagation when these bugs are reported.

Capture the noise the player generated, how it propagated, and which AI it reached, plus the alert state and how it spread, since these systems govern much of the stealth experience and their bugs break it. A report that a guard was alerted impossibly becomes clear when you can see the sound propagation that reached them or the alert state that spread to them. The sound and alert context, alongside the sight context, captures the full perception picture that stealth detection bugs require.

Setting it up with Bugnet

Add an in-game report option and attach the AI perception state, detection level, contributing senses, line-of-sight calculation, AI and player positions, and sound and alert state as custom fields, with a screenshot. Bugnet stores them so a stealth bug arrives with the AI-and-sightline context needed to reproduce a detection that misfired, rather than a subjective complaint about being spotted unfairly or ignored.

Enable automatic crash capture and group identical issues into occurrence counts, watching whether detection bugs cluster at particular locations, which would point at sightline geometry issues there. Because stealth detection is the fragile core of the genre and its bugs break the player trust in the rules, this perception-state capture is what lets you find and fix the detection bugs, restoring the consistent, fair detection that makes sneaking tense and satisfying rather than arbitrary and frustrating.

Test the detection edge cases

Stealth detection has many edge cases that produce bugs, partial cover, the player at the edge of a vision cone, sound at the threshold of detection range, multiple senses in conflict, and these are where detection most often misfires. Test these edge cases deliberately, probing the boundaries of the perception system, since that is exactly where the unfair-feeling detection bugs live, just as physics bugs live at the boundaries of the physics system.

Combine that edge-case testing with your captured detection bugs, which reveal the specific perception situations players hit that you did not anticipate. Your testing exercises the known boundaries of the detection system, and the captured reports surface the unexpected perception bugs from real play, the specific geometry, the particular sense combination, that produced an unfair detection. Together they let you tune the detection to be consistent and fair across the situations players actually encounter, which is the foundation of stealth that feels trustworthy rather than arbitrary.

Stealth lives on trusted detection. Capture the AI perception, the sightline, and the sound behind every unfair spot.